Urn!erf Slates      Office of A.r Quality        EPA--150 5-85-005&

             Tnvi.'ennent.i' Protection  Planning and Standards       August 1985

             Agency         Research Tnangle Park IMC 2771 1
I&EPA       Ambient Ozone And   M
             . .        . ,   .  .           ENVIRONMENTAt
             Human Health:
             An Epidemiological    DA'LAS'TEXAS
                               "^          MU^If*V
             Jt    I   •                     U* • Ski
             Analysis



             Volume  I

-------
                      AMBIENT  OZONE  AND HUMAN HEALTH:

                        AN EPIDEMIOLOGICAL  ANALYSIS
                     Paul  R.  Portney  and  John Mullahy
                         Resources  for  the  Future
                      1755 Massachusetts  Avenue, N.W.
                          Washington, B.C.   20036
                                 Volume I
                            Draft Final Report
                              September 1983
Submitted to the Economic  Analysis  Branch,  Office of Air  Quality  Planning
and  Standards,  Environmental  Protection  Agency,  Research Triangle  Park,
North Carolina 27711.  under contract number  68^02-3583.

-------
                                 DISCLAIMER









     This report has been reviewed by the Office of Air Quality Planning



and Standards, U. S. Environmental  Protection  Agency, and approved for



publication as received from Resources for the Future.   The analysis and



conclusions presented in this report are those of the authors and should



not  be interpreted  as necessarily reflecting  the official  policies of



the U. S. Environmental Protection Agency.

-------
                             ACKNOWLEDGEMENTS






     This report and the analysis on which it is based have benefitted from




the  efforts  of a  large number of  individuals.   Robert Fuchsberg  and his




colleagues in  the  Division  of Health Interview Statistics  at  the  National




Center  for  Health  Statistics  were  extraordinarily  helpful  in  making




available  to  us  and acquainting  us with  the intricacies  of the  Health




Interview  Survey.    Professor Raymond  Palmquist of  North Carolina  State




University  gave most  generously  of his  time  to  match  the  individuals




interviewed  in the  1979  Health  Interview  Survey  to  the  air  pollution




monitors nearest their homes.




     Within  the  Environmental Protection Agency,  we  received  support and




advice from  Allen  Basala, William  Hunt,  Bart  Ostro,  as well  as others  in




the Economic Analysis Branch,  the  Ambient Standards  Branch,  the Monitoring




and Data Analysis  Branch,  the Health Effects Research  Laboratory,  and the




Environmental  Criteria   Assessment  Office.    Thomas   Walton,  our  project




officer, not only  assisted  us  with countless administrative and procedural



details but also made valuable substantive suggestions at each stage of our




research.    V. Kerry   Smith,  Raymond  Palmquist,  Duncan  Holthausen,  Jan




Laarman,  and  W.   Kip  Viscusi—all  acting  as  consultants  to  EPA—also




provided valuable comments on our analysis.



     A number  of  people  knowledgeable  about air pollution  health effects




assisted  us  in  the early stages  of  our  research  by  identifying  air




pollutants that might plausibly  be associated with certain  types  of  acute

-------
and chronic illness.  We are most  grateful  in this  regard to Gilbert  Omenn,




Jane Koenig and Michael Morgan of  the Department  of Environmental  Health at




the University of Washington,  Michael  Lebowitz of  the College of  Medicine




at  the  University   of   Arizona,  and  Alice  Whittemore  of  the   Stanford




University School of Medicine.   John Spengler  of the Harvard University




School of  Public Health was  also very  helpful  in this  regard.   We  also




benefitted from  comments  received  during  seminars at  the University  of




California at  Berkeley,  Harvard University, Johns Hopkins  University,  the




University of Washington and Washington State University.




     Finally, we owe several of our  colleagues at  Resources for the  Future




a debt of  thanks.   Michael  Hazilla wrote the programs  which summarized the




mass of air quality  data with which  we worked and matched it to the periods




for  which had individuals'  health  histories.    Richard  Carson  provided




expert research assistance  in the  early stages of the  project.   Among those




who  commented  on early  versions  of  the  chapters,  Robert Frank  deserves




special mention.  He read  and commented extensively on both  the  form and




substance  of  every   chapter.   William  Vaughan,  Clifford  Russell, Winston




Harrington  and  Allen  Kneese  also  provided  very  helpful  comments  on




individual chapters.   Margaret Parr-Recard,  Pat Flynn, D.J.  Curran,  Anne




Farr  and  others worked  very  hard  typing  and helping produce this  draft




report.  We are grateful to them all  for their help.




     It  goes  without saying that none  of  the individuals  mentioned above




bears  any responsibility for  the contents  of  this report.   That  is  the




responsibility of the authors alone.

-------
                             "ABLE OF CONTENTS
VOLUME I
                                                                 Page
CHAPTER 1. INTRODUCTION
   1.1  General Background
   1.2  Economic Valuation and Physical Effects
   1.3  Identifying Air Pollution Health Effects
   1.4  The RFF Study
   1.5  Summary and Plan of this Report

CHAPTER 2.  DATA
   2.1  Overview
   2.2  The Health Interview Survey
        2.2.1  Main Survey
        2.2.2  Smoking Supplement
        2.2.3  Residential Mobility Supplement
        2.2.U  Documentation and Cross Referencing of HIS Data
        Air Pollution Data
   2.3
   2.5
        2.3.1  SAROAD System
        2.3.2  Data Integrity and Completeness
        2.3.3  Multi-Year Averaged Data
        2.3.4  Description of the Pollution Variables
        2.3.5  Matching Individuals to Monitors
        Meteorological Data
        Other Data
        2.5.1  Pollen
        2.5.2  Indoor Air Pollution
        2.5.3  Paid Sick Leave
        2.5.4  The Pollutant Standard Index
        2.5.5  availability of Medical Care
   2.6  Problems of Sample Selection

CHAPTER 3.  METHODOLOGY
   3.1  Measures of Health Status:
          Variables
   3.2  Explanatory Variables
   3.3  Estimation Techniques
                                    Choice of the Dependent
   Appendix to Chapter 3:
                           Definitions of and Procedures Used
                           to Create Measures of Morbidity
CHAPTER 4.  RESULTS AND DISCUSSION
   4.1  Acute Morbidity Due to All
        4.1.1  Adults
        4.1.2  Children
                                   Causes
                                                                  1-1
                                                                  1-4
                                                                  1-5
                                                                  1-15
                                                                  1-20
2-1
2-9
2-9
2-16
2-19
2-21
2-22
2-22
2-24
2-37
2-42
2-46
2-4?
2-51
2-52
2-54
2-56
2-57
2-59
2-60
                                                                  3-2
                                                                  3-6
                                                                  3-52
                                                                  A3-1
4-5
4-5
4-21

-------
   4.2  Acute Morbidity Due to Respiratory Disease                4-28
        4.2.1  Adults                                             4-31
        4.2.2  Children                                           4-55
   4.3  Chronic Respiratory disease                               4-59
        4.3.1  Adults                                             4-61
        4.3.2  Children                                           4-81
   4.4  Cardiovascular and Other Chronic Diseases                 4-85
   4.5  Analysis of Multicollinearity                             4-89

CHAPTER 5.  ALTERNATIVE OZONE CONCENTRATIONS AND CHANGES IN
            MORBIDITY:  ILLUSTRATIVE ESTIMATES
   5.1  Ozone-related Changes in Morbidity                        5-2
        5.1.1  Acute Morbidity                                    5-3
        5.1.2  Chronic Morbidity                                  5-6
   5.2  Acute and Chronic Illness:   Feedback Effects              5-8
   5.3  Relative Risks                                            5-11
   5.4  Monetary Benefit Estimates                                 5-18
VOLUME II

APPENDIX A.  The Health Interview and Smoking Supplement

APPENDIX B.  Health Interview Survey

APPENDIX C. -Valuing the Benefits of Improved Human Health

-------
                             TABLE OF CONTENTS
VOLUME I
                                                                 Page
CHAPTER 1. INTRODUCTION
   1.1  General Background
   1.2  Economic Valuation and Physical Effects
   1.5

CHAPTER
   2.1
   2.2
   2.3
     Identifying Air Pollution Health Effects
     The RFF Study
     Summary and Plan of this Report

     2.  DATA
     Overview
     The Health Interview Survey
     2.2.1  Main Survey
     2.2.2  Smoking Supplement
     2.2.3  Residential Mobility Supplement
     2.2.4  Documentation and Cross Referencing of HIS Data
     Air Pollution Data
            SAROAD System
            Data Integrity and Completeness
            Multi-Year Averaged Data
            Description of the Pollution Variables
            Matching Individuals to Monitors
            logical Data
            ata
            Pollen
            Indoor Air Pollution
            Paid Sick Leave
            The Pollutant Standard Index
            availability of Medical Care
2.6  Problems of Sample Selection





2.4
2.5





2.3.1
2.3.2
2.3.3
2.3.4
2.3.5
Meteoi
Other
2.5.1
2.5.2
2.5.3
2.5.4
2.5.5
CHAPTER 3.  METHODOLOGY
   3.1  Measures of Health Status:
          Variables
   3.2  Explanatory Variables
   3.3  Estimation Techniques
                                 Choice of the Dependent
   Appendix to Chapter 3'
                        Definitions of and Procedures Used
                        to Create Measures of Morbidity
CHAPTER 4.  RESULTS AND DISCUSSION
   4.1  Acute Morbidity Due to All Causes
        4.1.1  Adults
        4.1.2  Children
1-1
1-4
1-5
1-15
1-20
2-1
2-9
2-9
2-16
2-19
2-21
2-22
2-22
2-24
2-37
2-42
2-46
2-4?
2-51
2-52
2-54
2-56
2-57
2-59
2-60
                                                               3-2
                                                               3-6
                                                               3-52
                                                                  A3-1
                                                               4-5
                                                               4-5
                                                               4-21

-------
   4.2  Acute Morbidity Due to Respiratory Disease
        1.2.1  Adults
        4.2.2  Children
   4.3  Chronic Respiratory disease
        4.3.1  Adults
        4.3.2  Children
   4.4  Cardiovascular and Other Chronic Diseases
   4.5  Analysis of Multicollinearity

CHAPTER 5.  ALTERNATIVE OZONE CONCENTRATIONS AND CHANGES IN
            MORBIDITY:  ILLUSTRATIVE ESTIMATES
   5.1  Ozone-related Changes in Morbidity
        5.1.1  Acute Morbidity
        5.1.2  Chronic Morbidity
   5.2  Acute and Chronic Illness:   Feedback Effects
   5.3  Relative Risks
   5.4  Monetary Benefit Estimates
                                                      4-28
                                                      4-31
                                                      4-55
                                                      4-59
                                                      4-61
                                                      4-81
                                                      4-85
                                                      4-89
                                                      5-2
                                                      5-3
                                                      5-6
                                                      5-8
                                                      5-11
                                                      5-18
VOLUME II_

APPENDIX A.

APPENDIX B.

APPENDIX C.
 The Health Interview and Smoking Supplement

 Health Interview Survey

-Valuing the Benefits of Improved Human Health

-------
                             TABLE OF CONTENTS
VOLUME I
                                                                 Page
CHAPTER 1. INTRODUCTION
   1.1  General Background
   1.2  Economic Valuation and Physical Effects
   1.3  Identifying Air Pollution Health Effects
   1.4  The RFF Study
        Summary and Plan of this Report
   1.5
   2.1
   2.2
CHAPTER 2.  DATA
        Overview
        The Health Interview Survey
        2.2.1  Main Survey
        2.2.2  Smoking Supplement
        2.2.3  Residential Mobility Supplement
        2.2.U  Documentation and Cross Referencing of HIS Data
   2.3  Air Pollution Data
        2.3.1  SAROAD System
        2.3.2  Data Integrity and Completeness
        2.3.3  Multi-Year Averaged Data
        2.3.4  Description of the Pollution Variables
        2.3.5  Matching Individuals to Monitors
        Meteorological Data
        Other Data
        2.5.1  Pollen
        2.5.2  Indoor Air Pollution
        2.5.3  Paid Sick Leave
        2.5.4  The Pollutant Standard Index
        2.5.5  availability of Medical Care
   2.6  Problems of Sample Selection
   2.U
   2.5
CHAPTER 3-  METHODOLOGY
   3.1  Measures of Health Status:
          Variables
   3.2  Explanatory Variables
   3.3  Estimation Techniques
                                    Choice of the Dependent
   Appendix to Chapter 3:
                           Definitions of and Procedures Used
                           to Create Measures of Morbidity
CHAPTER 4.  RESULTS AND DISCUSSION
   4.1  Acute Morbidity Due to All Causes
        4,1.1  Adults
        4.1.2  Children
1-1
1-4
1-5
1-15
1-20
2-1
2-9
2-9
2-16
2-19
2-21
2-22
2-22
2-24
2-37
2-42
2-46
2-4?
2-51
2-52
2-54
2-56
2-57
2-59
2-60
                                                                  3-2
                                                                  3-6
                                                                  3-52
                                                                  A3-1
                                                                  4-5
                                                                  4-5
                                                                  4-21

-------
   4.2  Acute Morbidity Due to Respiratory Disease
        4.2.1  Adults
        4.2.2  Children
   4.3  Chronic Respiratory disease
        1.3.1  Adults
        U.S.2  Children
   4.4  Cardiovascular and Other Chronic Diseases
   4.5  Analysis of Multicollinearity

CHAPTER 5.  ALTERNATIVE OZONE CONCENTRATIONS AND CHANGES IN
            MORBIDITY:  ILLUSTRATIVE ESTIMATES
   5.1  Ozone-related Changes in Morbidity
        5.1.1  Acute Morbidity
        5.1.2  Chronic Morbidity
   5.2  Acute and Chronic Illness:   Feedback Effects
   5.3  Relative Risks
   5.4  Monetary Benefit Estimates
                                                      4-28
                                                      4-31
                                                      4-55
                                                      4-59
                                                      4-61
                                                      4-81
                                                      4-85
                                                      4-89
                                                      5-2
                                                      5-3
                                                      5-6
                                                      5-8
                                                      5-11
                                                      5-18
VOLUME II

APPENDIX A.

APPENDIX B.

APPENDIX C.
 The Health Interview and Smoking Supplement

 Health Interview Survey

-Valuing the Benefits of Improved Human  Health

-------
                             TABLE OF CONTENTS


                                                                 Page
VOLUME I
CHAPTER 1. INTRODUCTION
   1.1  General Background                                        1-1
   1.2  Economic Valuation and Physical Effects                   1-4
   1.3  Identifying Air Pollution Health Effects                  1-5
   1."  The RFF Study                                             1-15
   1.5  Summary and Plan of this Report                           1-20

CHAPTER 2.  DATA
   2.1  Overview                                                  2-1
   2.2  The Health Interview Survey                               2-9
        2.2.1  Main Survey                                        2-9
        2.2.2  Smoking Supplement                                 2-16
        2.2.3  Residential Mobility Supplement                    2-19
        2.2.4  Documentation and Cross Referencing of HIS Data    2-21
   2.3  Air Pollution Data                                        2-22
        2.3.1  SAROAD System                                      2-22
        2.3.2  Data Integrity and Completeness                    2-24
        2.3-3  Multi-Year Averaged Data                           2-37
        2.3.4  Description of the Pollution Variables             2-42
        2.3-5  Matching Individuals to Monitors                   2-46
   2.4  Meteorological Data                                       2-47
   2.5  Other Data                        "                        2-51
        2.5.1  Pollen                                             2-52
        2.5.2  Indoor Air Pollution                               2-54
        2.5.3  Paid Sick Leave                                    2-56
        2.5.4  The Pollutant Standard Index                       2-57
        2.5.5  availability of Medical Care                       2-59
   2.6  Problems of Sample Selection                              2-60

CHAPTER 3.  METHODOLOGY
   3.1  Measures of Health Status:  Choice of the Dependent
          Variables                                               3-2
   3.2  Explanatory Variables                                     3-6
   3.3  Estimation Techniques                                     3-52

   Appendix to Chapter 3:  Definitions of and Procedures Used
                           to Create Measures of Morbidity        A3-1

CHAPTER 4.  RESULTS AND DISCUSSION
   4.1  Acute Morbidity Due to All Causes                         4-5
        4.1.1  Adults                                             4-5
        4.1.2  Children                                           4-21

-------
   4.2  Acute Morbidity Due to Respiratory Disease
        4.2.1  Adults
        4.2.2  Children
   4.3  Chronic Respiratory disease
        4.3.1  Adults
        U.S.2  Children
   4.4  Cardiovascular and Other Chronic Diseases
   4.5  Analysis of Multicollinearity

CHAPTER 5.  ALTERNATIVE OZONE CONCENTRATIONS AND CHANGES IN
            MORBIDITY:  ILLUSTRATIVE ESTIMATES
   5.1  Ozone-related Changes in Morbidity
        5.1.1  Acute Morbidity
        5.1.2  Chronic Morbidity
   5.2  Acute and Chronic Illness:   Feedback Effects
   5.3  Relative Risks
   5.4  Monetary Benefit Estimates
4-28
4-31
4-55
4-59
4-61
4-81
4-85
4-89
5-2
5-3
5-6
5-8
5-11
5-18
VOLUME II_

APPENDIX A.  The Health Interview and Smoking Supplement

APPENDIX B.  Health Interview Survey

APPENDIX C. -Valuing the Benefits of Improved Human Health

-------
                                   OF CONTENTS
VOLUME I
                                                                 Page
CHAPTER 1. INTRODUCTION
   1.1  General Background
   1.2  Economic Valuation and Physical Effects
   1.3  Identifying Air Pollution Health Effects
   1.14  The RFF Study
   1.5  Summary and Plan of this Report

CHAPTER 2.  DATA
   2.1  Overview
   2.2  The Health Interview Survey
        2.2.1  Main Survey
        2.2.2  Smoking Supplement
        2.2.3  Residential Mobility Supplement
        2.2.4  Documentation and Cross Referencing of HIS Data
   2.3  Air Pollution Data
               SAROAD System
               Data Integrity and Completeness
               Multi-Year Averaged Data
               Description.of the Pollution Variables
               Matching Individuals to Monitors
              Dlogical Data
              )ata
               Pollen
               Indoor Air Pollution
               Paid Sick Leave
               The Pollutant Standard Index
               availability of Medical Care
   2.6  Problems of Sample Selection

CHAPTER 3.  METHODOLOGY
   3.1  Measures of Health Status:
          Variables
   3.2  Explanatory Variables
   3.3  Estimation Techniques




2.4
2.5





2.3-1
2.3.2
2.3.3
2.3.5
Meteoi
Other
2.5.1
2.5.2
2.5.3
2.5.4
2.5.5
         Choice of the Dependent
   Appendix to Chapter 3:
Definitions of and Procedures Used
to Create Measures of Morbidity
CHAPTER 4.  RESULTS AND DISCUSSION
   4.1  Acute Morbidity Due to All Causes
        4.1.1  Adults
        4.1.2  Children
                                       1-1
                                       1-4
                                       1-5
                                       1-15
                                       1-20
                                       2-1
                                       2-9
                                       2-9
                                       2-16
                                       2-19
                                       2-21
                                       2-22
                                       2-22
                                       2-24
                                       2-3?
                                       2-42
                                       2-46
                                       2-47
                                       2-51
                                       2-52
                                       2-54
                                       2-56
                                       2-57
                                       2-59
                                       2-60
                                       3-2
                                       3-6
                                       3-52
                                                                  A3-1
                                       4-5
                                       4-5
                                       4-21

-------
   4.2  Acute Morbidity Due to Respiratory Disease                4-28
        I*. 2.1  Adults                                             4-31
        4.2.2  Children                                           4-55
   4.3  Chronic Respiratory disease                               4-59
        4.3.1  Adults                                             4-61
        4.3.2  Children                                           4-81
   4.4  Cardiovascular and Other Chronic Diseases                 4-85
   4.5  Analysis of Multicollinearity                             4-89

CHAPTER 5.  ALTERNATIVE OZONE CONCENTRATIONS AND CHANGES IN
            MORBIDITY:  ILLUSTRATIVE ESTIMATES
   5.1  Ozone-related Changes in Morbidity                        5-2
        5.1.1  Acute Morbidity                                    5-3
        5.1.2  Chronic Morbidity                                  5-6
   5.2  Acute and Chronic Illness:   Feedback Effects              5-8
   5.3  Relative Risks                                            5-11
   5.4  Monetary Benefit Estimates                                 5-18
VOLUME II_

APPENDIX A.  The Health Interview and Smoking Supplement

APPENDIX B.  Health Interview Survey

APPENDIX C. -Valuing the Benefits of Improved Human Health

-------
                                 Chapter 1
                               INTRODUCTION
\. 1  General Background




This report  presents  the  methodology and results  of  a study undertaken by




Resources  for  the  Future  (RFF)  for the Economic Analysis Branch  of the



Environmental  Protection  Agency's  Office   of  Air  Quality Planning and




Standards  (OAQPS),  The study  was designed  to identify the benefits in the




form  of  improved  human health associated  with  possible  alternative air




quality standards for ozone.




     One act  of  Congress and  one executive  order provide  the  impetus for




this study.  In  1970,  Congress amended  the  Clean Air Act of 1963 in such a




way as to fundamentally alter the approach by which air quality was pursued




in the United States.  Among other things, the 1970 Amendments to the  Clean




Air Act  directed the Administrator  of  the  Environmental  Protection Agency




(EPA)  to establish  two  kinds  of National  Ambient Air  Quality Standards




(NAAQS)  for  a handful of  common (or "criteria")  air pollutants.  Primary




standards  were  those  designed  to protect  human health  with  an "adequate




margin  of   safety."    Secondary  standards  were  also  to be  issued  where




appropriate; these  were  to be  set  at  levels  which afforded protection to

-------
                                    1-2






vegetation,   aquatic   ecosystems,   visibility,   and   other   "welfare"




considerations (where "welfare" refers  to  non-health concerns, rather than




to general well-being as in economists' parlance).




     Accordingly,  in  1971   EPA  promulgated  NAAQSs   for  total  suspended




particulates,   sulfur  dioxide,  carbon  monoxide,  nitrogen  dioxide,  and




photochemical   oxidants.   A  standard  was  also  issued  for  hydrocarbons




although it was  intended only  to  assist states  and  localities  in meeting




the NAAQS for photochemical oxidants.   In  1978  a NAAQS was also issued for




airborne concentrations  of lead.   To this date,  these same six pollutants




are the only ones governed by NAAQSs.




     Although   the  1970 amendments  to  the   Clean  Air  Act   directed  the




administrator  of  EPA to revise  the  ambient standards  whenever  new data




accumulated  to  so  warrant,  in  1977  Congress  provided  more  specific




guidance.  The  1977 amendments directed EPA  to  review each NAAQS at least




every five years and to revise the existing standard where it seemed called




for, or  retain  the existing  standard  if no   compelling  evidence suggested




that it  should  be changed.    In  1978, in  the  first of  these mandatory




revisions, EPA  proposed to change  the  basis  of  the  photochemical oxidant




standard to ozone (the most common photochemical oxidant), and to relax the




standard from  .08 parts-per-million (ppm), not  to be exceeded  during the




highest  hour  each  day  more  than  once per  year,  to .10  ppm  not  to  be




exceeded  (on  an  expected value basis)  more  than three times  over a three




year period.  In  1979,  EPA issued  a final  revised NAAQS for ozone.  It was




set  at  .12  ppm,  to  be  measured  in  the way  suggested  in  the  proposed



standard.

-------
                                    1-3






     Because of the five-year mandatory review schedule established in the




1977 amendments to  the  Clean Air Act,  EPA  is  now reviewing the scientific




evidence relating  ozone to  human  health,  agricultural  output, visibility




impairment,  and  other  effects.   One  purpose  of  the  present  study  is  to




provide  some  additional information  to the EPA,  and specifically  to the




OAQPS,  as part of that review.




     However, the  study reported here was  also  undertaken on  account  of




Executive  Order  12291,  issued  by  President Reagan  on February  17,   1981.




That order, a  successor  to Executive Order  12044  issued by  President




Carter,  directed all  federal regulatory agencies  in  the  executive branch




to prepare  both preliminary  and  final  Regulatory  Impact Analyses (RIAs)  to




accompany  all  proposed and  final  "major"  federal rules  and  regulations.




These  RIAs are, among  other things,  to  describe the potential  benefits




expected to result  from proposed  or  final  rules, identify  the potential




costs,   calculate  the  net  benefits   (the  former  minus  the latter),  and




describe   alternative approaches that might accomplish the  same things  as




the rule in question,  but at a lower cost—along with an explanation of the




legal reasons why such alternatives  could not be selected if, in fact, they



were not.




     Because  the  promulgation  of NAAQSs  have  been  designated as  major




rulemakings, Regulatory Impact  Analyses  have  generally  accompanied  them.



Thus, another purpose  of the present  study is to provide information on the




human health benefits  associated with possible alternative NAAQSs for ozone




so  as   to   facilitate  the  compilation  of  the  overall  benefits—health,




agricultural,  materials  damage,  aesthetic,  and  other—associated  with

-------
                                    1-U






different ozone standards.  Such information,  compiled  from a wide variety




of sources, could  then  be combined with data  on  the  costs associated with




these  same alternatives  to  provide   information on  the  overall  effects




expected to result from varying possible ozone standards.








1.2  Economic Valuation and Physical Effects




The original purpose  of  this  study was to  estimate,  in dollar terms where




possible,  the  value of  any reduced human  morbidity that  might accompany




reductions  in  ambient ozone  concentrations  (or  the additional morbidity




that  might  arise  from  a relaxation of   the   standard).    We  were  to




concentrate on both acute as  well  as  chronic  morbidity.   That is,  we were




to concern ourselves with the  temporary or day-to-day changes in individual




health status  that  might  result from human  exposure  to ambient ozone, and




how these might be valued.  In addition, we were to attempt to place dollar




values on any long-term, or chronic, illness that might result, perhaps not




so much from occasional exposures  to peak  ozone  concentrations, but rather




from prolonged exposures to ozone as well as other air pollutants.



      Although the  overall  focus of the  work reported  below  is still the



same, the  emphasis has  shifted somewhat during  the course of  our study.




Specifically, because of  the  availability of an  extraordinary data set, it




became clear  during the  initial phase of  the RFF study  that  we might not




only  say  something original  about  the economic .valuation  of  ozone health




effects,  but  might  also  be   able   to   provide  information  about  the




relationship between  ambient  ozone  concentrations and  the health effects




themselves.  The  question of economic  valuation  has  received considerable

-------
                                    1-5






thought during the course of our study.  Appendix C to this report presents




a  model  of  household  behavior from  which  are derived  measures  of  the




valuation of health  improvements.   However, more work  than was originally




anticipated was devoted to identifying the acute and chronic health effects




that  might  be  related  to  ambient   exposures  to  ozone  and  other  air




pollutants.  Thus, this report concentrates primarily on the identification




and quantification of health effects resulting from exposure to ozone.








1.3  Identifying Air Pollution Health Effects




There  are  three  primary means  of ascertaining the relationship between




ozone  or  other air  pollutants  and both  acute  and chronic morbidity.   In




this  report  we  shall  refer  to  these  three  approaches  as toxicological,




clinical, and  epidemiological.   Although  we  make  use of  but  one of these




approaches in our study, it is useful to discuss all three albeit briefly.




     Under the toxicological approach, laboratory animals are exposed under




carefully  controlled   conditions   to  varying  levels   of  specific  air




pollutants.  Dose-response  relationships  are estimated  either by exposing




otherwise identical  groups of  animals to different air pollution regimes




and  then observing  differences  in  biological  end-points  (which  may be




functional,  biochemical,  structural,  or  behavioral),  or by  exposing  the




same  animal(s)  to  alternative  ozone  or other  air pollution  levels at




different  times  and  observing  changes  in  the   same  end-points  under




different regimes.  The information gleaned from such  animal tests is  used




to provide insight  into the effects of these same  air  pollutants on human




health.

-------
                                    1-6






     As with the  other  approaches  to health effects  estimation,  there are




both advantages  and disadvantages  to  the  toxicological  method.    On the




positive  side,  animals  can   be  subjected  to  carefully  controlled  air




pollution levels, at least  some of  which experimenters  would  be  unwilling




to administer to human  subjects.  In  addition,  since  the lifetimes of some




laboratory animals,  as  for example mice, may be only about two years, it is




possible to observe  the effects of  a simulated  "lifetime"  exposure to air




pollution in a manageable period of  time.   Finally, laboratory animals can




be  sacrificed  during the  course of  their lives  to  observe  directly the




timing  of  physiological  changes  resulting  from  the  exposure  to  air




pollution that  may occur anywhere within the organism.




        There  are,   however,   several  serious  drawbacks  to  toxicological




studies as a means  of  determining human susceptibility.   The  most obvious




of  these  involves  the  uncertainty   of  extrapolating the  results  seen in




animals  to  human  beings.-    Researchers are  unsure  whether  humans  will




respond  even  qualitatively  to  air  pollution  (or  other  environmental




stimuli) in the same way  as do experimental  animals,  much less whether the




quantitative dose-response  relationships estimated in  animal  studies are



representative  of  likely human  responses.    In  addition,   the   cost  of




toxicological studies sometimes  necessitates  that  animals be given "doses"




of  air  pollution much  greater than levels  to  which  humans might  ever be




exposed to make sure an effect is observed if it exists.  Thus, there often




arises  the  additional  problem of extrapolating from effects  observed at




high  doses  to  those   observed  at   much  lower doses,  even  before  the




inter-species extrapolations must be made.

-------
                                    1-7


     A second approach to human  health effects  estimation is the clinical.

At   the   risk    of   over-simplification,   this   involves   controlled

experimentation  in exposure  chambers  or  other laboratory  settings using

human subjects.  Typically, volunteer subjects are exposed acutely—usually

for periods lasting  up to  a few  hours—to a variety of different levels of

air  pollution,   perhaps  under different  exercise  or  other  conditions as

well.   Their  symptomatic,  physiological, or  in some instances biochemical

responses  are monitored  to  provide  evidence  about  the  effects   of  air

pollution on human health.

     There are several desirable features  of  clinical  studies.  First, and

perhaps foremost,  the dose  of  any  pollutant(s)  can be carefully controlled

and measured.   Thus, we  can be quite  sure that a  certain  individual was

exposed to, say,  .20 ppm  ozone for exactly two hours and in the absence of
                                                                     •
other  pollutants,  because  the air inflow  to a chamber  can be controlled

quite  closely.   Also, because individuals  have the opportunity to report

symptoms as they are experienced,  or  because physiological  changes can be

measured  in  "real  time"   (i.e.,   contemporaneously),  the  likelihood  of

establishing  a  causal link is quite  good.   In other  words, establishing

such a  link does  not depend on  an individual's recollection  of a  symptom

sometime in  the  recent  or  distant past.  Finally,  if  the  pollutant(s) in

question   are   thought   to  be   particularly   harmful   to   "sensitive

individuals"—those  who  because  of age, underlying  disease,  or some other

distinguishing  characteristic, are  thought  to be  particularly adversely

affected by  the pollutant—these  individuals can be selected for  study.'

In other words,  the experimenter  need not rely on  good  fortune to ensure

-------
                                    1-8






that there are  enough  children,  asthmatics,  or other sensitive individuals




in the sample group.




     In spite of  these attractive aspects, there  are  several drawbacks to




the  clinical  approach.    First,   and  least  serious,  concerns  the  air




pollution dose  that  subjects  receive.  Although it  is  generally carefully




measured, it is also artificial  and typically  is  not representative of the




entire mixture  of air  pollutants  and  other substances  to which individuals




are normally exposed in the course of  their everyday  lives.  For example,




it may not contain  any of  the  allergens or infectious  material to which we




are all exposed from time to time.  Thus if there are important synergistic




effects  between several  pollutants,  they may be overlooked  in clinical



studies.




     Another  problem   with clinical  studies  concerns  the  numbers  of




participants.     Perhaps   on  account   of  cost,   clinical  studies  are




generally—although not always—characterized by small  numbers of subjects.




It is  not  uncommon  to  find studies with  fewer than te.i subjects, while it




is unusual to see studies in which more than fifty individuals participate.




This creates  two  problems.  First, it limits  one's ability to generalize




the  results  of  a  particular  study  to larger,  potentially  more  diverse




populations.    Such  generalizations  will  be  especially difficult  if the




individuals  studied are more  sensitive in  some  respect than the  overall




population.



     Small  samples  create  another  problem.    For  any   given  level  of




statistical  significance,  the  smaller  the number  of experimental subjects




the larger must be  any observed  effect.  Since at least some  air pollution

-------
                                    1-9






health effects  can  be  quite subtle, this  means  that potentially important




relationships can sometimes  be missed.   Larger sample sizes make this less




likely.



     These problems pale  in  comparison  to the most serious drawback of the




clinical  approach—its  general  inability  to shed  light on  the possible




long-term  or  chronic health effects  associated with  ozone  or  other air




pollutants.  For  while  subjects  can be placed in artificial settings where




exposures  are  controlled  for hours  or  perhaps  even  days or  weeks, legal,




ethical and practical considerations prevent  use of exposure regimes likely




to  cause  persistent or irreversible  damage,  or that would  take years to




complete.  Yet  one  important concern  about air pollution is that prolonged



exposures,  even  to  levels  well  below  those at  which  acute  effects are




thought  to begin, may  eventually result  in chronic respiratory or other




kinds of  disease.   Clinical  studies cannot resolve this  concern.  In fact,




air  quality standards  designed  primarily  to  guard against  acute  health




impairments associated with  air  pollution might  still permit  the onset and




exacerbation of chronic illnesses  that  ought also to be  taken into account




in  setting health-based air quality  standards.   It  is  here  that reliance



must be placed on epidemiological research.




     An  epidemiological  study  is  one  that  looks  at different  groups of




individuals at one point in  time or the same individuals  at different times




(or both)  and  attempts  to correlate differences  in their health status to




other differences between them  (environmental,  personal, and  other).    Of




course,  health  status  is   affected  by  much more  than  exposure  to air



pollution  (or even  all  environmental  factors) and  these  other factors can

-------
                                   1-10






vary dramatically across different individuals  or  population groups at any




one time;  they  can  also differ for specific  individuals  over time—income




may  grow,  personal  habits   (particularly smoking)  may  change,  weather




varies, marital  status,  occupation,  education and a host  of other factors




may  change  and  all  may  significantly  influence  health.    A  goal  of




epidemiological  research  is   to  hold  constant  as  many  of  these  other




identifiable  influences  as  possible.    This  may  be  done  either  through




statistical  means  in  the  case  of  a  heterogeneous  population,  or  by




deliberately selecting sample populations that are alike in as many ways as




possible  save  for   the  characteristic(s)  whose  potential  influences  on




health is to be isolated.




     Epidemiological studies  can be distinguished in a number of ways.  One




distinction revolves  around  the units of  observation,  whether individuals




or aggregate  populations.    In the former, data are  available on specific




individuals—their    health     status,     socioeconomic    characteristics,




environmental  exposures,  and  so  on.   In  aggregate  epidemiology,  on the




other hand, a researcher tries to link variations in morbidity or mortality




rates  for different  cities,  counties,  states  or even  countries  to air



pollution  after   attempting  to  control  for  the   variation   in  other




potentially important  explanatory variables.     Individual  data  are always




to be  preferred where it is  available because  it  enables  one to avoid the




so-called "ecological fallacy" which can arise when averaged or pooled data




obscure important individual  characteristics.




     Another  distinction  that  can  be  drawn  among epidemiological studies




concerns the extent to which the study is controlled or uncontrolled.  In a

-------
                                   1-11






controlled (or ex ante)  study,  the  individuals or groups to be studied are




identified ahead of  time.   They are  then  monitored  over the course of the




study period for changes in health status, personal characteristics, and so




on.    In  addition,  ozone  concentrations,  say,  are  carefully  monitored




during the  study period so  that  they can  be  linked to  health  status (or




other dimensions of  interest).  In  other words, such controlled or ex ante




studies are not  too  unlike  clinical  studies save for the fact that the air




pollution exposures take place in real rather than artificial settings, and




for the possibly extended duration of epidemiological studies.




     In uncontrolled  (or ex post) epidemiological  studies, the researcher




may  examine  the health status of  individuals  or  groups  for  whom health



status, other information, and  air  pollution exposures  are known or can be



constructed,  even  though  some  or  all  of  this information may  have  been




collected for reasons  completely unrelated to  air pollution or other kinds




of  epidemiology.'     In either  controlled  or  uncontrolled  studies,  the




quality of the  results depend critically  upon  the  breadth and reliability




of the  data  on  both  the variables  of interest  (exposure to air pollution




and health status)  as well as the potentially confounding influences.



     There  are  a  number  of  desirable  features  to  the  epidemiological



approach.    Foremost  among  these  is  that,  in  principle,  epidemiological



studies have  the  capability of  shedding  light on  any  possible  chronic




health  effects  associated  with air  pollution or other factors.   If one




could  identify  individuals  who have lived  in  one  location  their  whole




lives, if one had reliable measures of their health status, if one knew the




levels of  ozone  and  other  air  pollutants to which  they were  exposed over

-------
                                   1-12




this  period,  and if  one could  control  for the  many other  influences  on



health,  then  one  could draw  some  (cautious)  inferences  about the  link



between air pollution and health.  Obviously, these  are very big "ifs" and



the likelihood of their  being satisfied  is  discussed below.  Nevertheless,



we emphasize  again  that, unlike clinical studies, epidemiological  studies



can  in  principle  make  an  important  contribution  to  the  study  of  air



pollution and chronic illness.



     Another advantage to epidemiological studies is  that, in general, they



include  larger  numbers  of  individuals  than  typically  participate  in



clincical studies.  This is not inherent in the nature of such studies, but
            >


it is  generally the  case.   Such large  numbers  make it more likely that any


statistically significant relationships will be identified—be they between



air pollution and health status  or between  other dependent and independent



variables.



     There is a final, less important advantage to epidemiological studies.



By definition, such  studies  are  based  on ambient or  "real world" exposures



to  air  pollution  and  other environmental  factors,  rather  than  to the



artificial  exposures  administered   in   clinical  studies.    Thus,  if the



exposure   data   are   measured   correctly,   they   accurately   (indeed



definitionally) reflect  the  conditions  under which people actually receive



doses  of pollution.   Such  studies  avoid the  problems  that may arise  if



laboratory   settings   induce    by  themselves   certain  behavioral  or



physiological changes.



     Like   the   toxicological   and   clinical    approaches,   however,  the



epidemiological approach is  not without problems.   It,  too,  has drawbacks

-------
                                   1-13




that can  seriously limit  the  confidence one  can  place in the  results  of


epidemiological  studies.     The  most  serious  of   these   concerns  the


measurement of  individual  exposures to  any  particular air pollutant.   In


contrast  to  clinical  studies,  where  concentrations  can  be  carefully


controlled, epidemiological  studies typically rely  for  exposure  data  on


readings from ambient air pollution monitors in the areas where individuals


live, work and recreate.  In some cases, in fact, no monitoring data may be


available  from  the area in which  one  or more individuals live.   And even


where data are available for a particular region, they may be suspect for a


number  of  reasons.   Since individuals  may  in  the  course of  their daily


activities travel all about a particular metropolitan area, one must decide


whether to characterize air  pollution dose  by  the  monitor  nearest their


homes or  offices,  or take some  kind of  weighted average based on relative


amounts of time spent in various locations.


     Often no  data whatsoever are available  about  such activity patterns,


so  that even  in the presence of comprehensive ambient monitoring data,  it


would be very difficult  to create  a weighted exposure index.   Complicating


the  estimate  of  exposure  still further  is   the matter  of  indoor versus

                                    Q
outdoor exposure  to air  pollution.    It  is  now known, for  example, that


individuals may be exposed to much higher peak  concentrations of nitrogen


dioxide, particulate matter, and carbon  monoxide in indoor than in outdoor


environments  even  though  these  are  three   pollutants  for  which  ambient


standards  exist.     In  addition,  individuals  may  be  exposed to  radon,


formaldehyde,   carbon   dioxide  and  other   air  pollutants   in   indoor


environments,  and  these  indoor concentrations  may  make  it  difficult  to

-------
                                   1-14


isolate the effects on health of ambient  air  quality.   (Because this study

deals  specifically  with  the health  effects associated  with ozone,  it  is

noteworthy  that  ozone  is   one  of   the  air  pollutants  that  is  almost

non-existent  in  indoor   environments.    With  the  exception  of  certain

occupational settings  (for  example,  welding), almost  all  exposures result

from contact with ambient air.)

     The seriousness  of  the problem of estimating exposure  varies between

epidemiological  studies.    In   some   studies  the  nearest  air  pollution

monitors are so far removed from the subjects, or their data so unreliable,

that little  confidence can  be  placed in the  results.  In  other studies,

however, monitoring is done nearby  and  very  carefully.  In  these cases,

exposures tcay  be well represented.   Nevertheless, ambient  monitoring data

can  be  no  more than a proxy for actual  total exposure  to ozone  and other
                                          •
air pollutants.

     Another problem, about which less will be  said,  concerns  the controls

in   epidemiological   studies  for  variables   other   than   air  pollution.

Clearly, the greater the  extent  of  controls for such variables, the greater

the  confidence  that can  be  placed  in  any air pollution health effect that

is  identified.    The  extent of these  controls varies markedly between

studies—in  some,  such obvious  and  potentially confounding influences  as

personal smoking habits,  pre-existing respiratory disease,  or  exposures to

other  air  pollutants  are missing.    In other studies,  earnest efforts are

made to account  for these and other  much more subtle influences on health.

As in the case of exposure data in epidemiological studies, the severity of

-------
                                   1-15






the "control problem" depends on the specific measures  taken in each study




to address the problem.








1.U  The RFF Study




The  study  reported  here  is an  epidemiological one.    It  combines  four




different types of data to  shed light  on acute  and  chronic human morbidity




that may  be associated  with ozone, other criteria  (and some non-criteria)




air pollutants, and other possible contributory factors.




     As described in detail  below, the  data  on  acute and chronic morbidity




as well as socioeconomic characteristics are specific to individuals rather




than  to  population  groups.    Thus,   ours  is  an  individual   (or  micro-)



epidemiological study.   The  individual  health and  socioeconomic  data for




this study  come from the  1979  Health  Interview Survey  (HIS)  of  more than




110,000 individuals, a  survey that is  conducted annually by  the National




Center for  Health  Statistics (NCHS)  in the Department  of Health and Human




Services  (HHS).  The survey is  reproduced in its entirety in Appendix A to




this report.




     The  health data are  of two  types.   First, information  about acute




illness was elicited from  all  individuals surveyed, information  that was



specific to the two-week period ending the  Sunday night before the week of




the  interview.   Individuals were asked about  physical conditions  that




confined them  to bed for most of  a day, prevented normal activity (work or



school) without requiring  confinement  to  bed,  or forced  them to restrict




their normal activity in some way  even though it was not serious enough to




force them  to  stay in bed  or miss work or school.   Individuals  were also

-------
                                   1-16






asked about doctor visits during the two-week  period  as  well as short-stay




hospital episodes, although these are not analyzed in our report.




     In addition to information about acute health status during a specific




two week period, invididuals participating in  the  1979 HIS were also asked




whether  they  had  any  chronic  health  impairments  that  limited  their




activities.   Adult  respondents  could also  identify chronic  illnesses  or




impairments through their responses to a  separate  section of the 1979 HIS.




Both acute  and  chronic impairments were  identified by  cause  and recorded




with the International Classification of Disease (ICD) codes.




     One  of  the  advantageous  features   of  the  HIS  is the  form  of  the




questions about acute health status.  Because such concepts as a. "work loss




day" or  a  "bed  disability day" are easily understandable,  individuals may




be  able  to  value   the   prevention   of   such  occurrences  without  much




difficulty.   This  is important in a study like  this  one—one aim of which




is to translate physical  improvements  in health into dollar terms—because




these  individual   valuations   are  the  basis  of  benefit  calculations  in




welfare economics.



     This would  be difficult  to  accomplish  if each  of  the respondents in




the HIS  were  given a test of  lung function as part of the examination.  In




such a  case, we  would in our  statistical  analysis be  identifying a link




between  ozone concentrations and lung  function.   If we  calculated a change




in the  latter for a given change  in  ozone,  the resulting effect—although




important to  a  clinical  appraisal—would  not  lend itself  to  valuation as




easily as a practical endpoint like one extra or fewer day spent sick in

-------
                                   1-17






bed.   Thus,  from  the  standpoint  of applied  welfare  economics  the  HIS




elicits health information in a useful form.




     The  socioeconomic  data  collected  for  each  individual   in  the  HIS




include  age,  race,  sex,   income,   education,   occupation,   industry  of




employment, marital status, history of military service, as well as




other individual and household-specific characteristics discussed below.




     In  addition  to  the  health  and  socioeconomic  information  that  is




elicited  each  year as  part of  the  HIS, several  supplements  to  the  1979




survey   made  it   particularly  useful   for   epidemiological   purposes.




Specifically, the 1979 HIS contained a supplement that went to one-third of




all  the  adults  interviewed  (26,271  of a  total  of  79,743 adults)  which



provided detailed  data  on lifetime  smoking history, including  the tar and




nicotine  content  of   the  brands  most  commonly  smoked.    (The  smoking




supplement is included in Appendix A.)   This  is  of great importance if one



is interested in examining respiratory and cardiovascular disease.




     Also, the  1979 HIS included a  supplement   (again  to  one-third  of all




adults surveyed)  designed to  provide detailed  information on residential




histories.   This  information is important in examining the  possible links




between  air  pollution and chronic respiratory  disease, since  it  makes it



possible to  identify  individuals who have lived in  their  present  location




for  a  long period  of time.   Supplemental questions were  also asked about




eye care and home health care in the 1979 HIS.




     To  this rich  set  of  individual health,  socioeconomic,  smoking and




residential  history data,  RFF added  three  other types of data.   The most




important  of these  were data on air  quality  in  the  United States  in 1979»

-------
                                   1-18






consisting of  all the  hourly  or 24-hour  readings  from the  air pollution




monitors  that  reported  to EPA's  SAROAD  file  (JStorage  a.nd Retrieval  of




Aerometric Data)  in  1979, as  well  as summary measures  of this  data.   We




also merged with  the individual HIS  data, other annual average pollution




data dating  back  in some cases as  far  as  197**  (again from EPA's SAROAD




network). Thus, we used air pollution data  that  were  contemporaneous with




the period for which we had information on individuals' acute morbidity (as




opposed  to  using  annual  average air  pollution to  proxy  exposure during a




specific  part  of  the  year).    In  addition, we had  annual  average  air




pollution data from  the  urban  areas  in which the  individuals in our sample




resided  (for both  ^979 as well as previous  years in many cases) to assist



in  exploring  the  possible role  of air  pollution  in  chronic  disease




etiology.




     In  addition  to this contemporaneous matching  of individuals  to  air




quality  data,  we  also  tried for a finer  spatial  resolution of  air quality




data.     Using  a  procedure  discussed   in  Chapter  2, we  matched  those




individuals  in the  1979  Health  Interview  Survey  who  lived  in Standard



Metropolitan Statistical Areas  (SMSAs) to the nearest ten monitors for each




of  eight pollutants—ozone, total  suspended particulate  matter, nitrogen




dioxide,  total oxides   of  nitrogen,  sulfur  dioxide, carbon  monoxide, lead




and  sulfates.   In addition, we knew the  distance  between the  centroid of




the  census  tract  in  which an  individual  resided  and the location of these




monitors.    Thus,  although  we  were  free  to  use  SMSA-wide  averages  to




represent air  pollution exposures  at  any point   in  time,  or construct our




own  distance-weighted  average  exposures,  we had the capability to match

-------
                                   1-19






each individual in  our  sample to the air  pollution  monitor(s) nearest his




or her  home and  use  that as  a  measure of  exposure.   The latter  was the




approach used in our study.




     We  also added  to  our  data  base  information  on weather  conditions




specific to  the  two-week period for  which we had acute health  data for a




particular individual, as well as  averaged over the  year  1979.  These data




came from  the National  Oceanographic  and  Atmospheric Administration (NOAA)




and were measured at  one or more  weather  stations  in each SMSA, generally




at airports.  Included in the weather data were observations on temperature




(both central  tendencies  and extremes),  precipitation, humidity and other




measures.




     Finally, we  compiled a  third data  set,  one much more  heterogeneous




than the  air pollution or meteorlogical  data.  This  included  a number of




potential  influences  on  acute  and  chronic  health  status  which  were




unrelated to socioeconomic status, air pollution concentrations, or weather




conditions.  Included in  this  category  were:   ragweed concentrations for a



subset of  the metropolitan  areas from which  the  individuals  in our sample




were drawn;  the probability that  an  individual came from a household where




piped or  bottled  natural  gas  was  the predominant  cooking  fuel (because of




its potential for indoor air pollution);   a  measure  of the availability of




doctors in the area in which an individual lived; an estimate of the annual




amount of paid sick leave to which an individual was entitled  (because sick



leave may affect  work  loss);  as well  as  a number of  other  unrelated but




potentially  important factors.

-------
                                   1-20






     A  final  observation  about  the  RFF  study  concerns  our reliance  on




medical and epidemiological experts.   Every  study  should try to build upon




and learn from the advances made  in previous  studies.  The present study is




no  exception.   We  have tried to draw  on the  best features of  previous



epidemiological, clinical and toxicological  studies.   These were useful in




helping  identify   possible   explanatory  variables   and  in  constructing




hypotheses about the kinds of acute and chronic health impairments expected




to result from  short-term peak exposures  or  longer-term exposures to ozone




and  other air  pollutants.   Of  particular  assistance were  a number  of




doctors  and/or   epidemiologists   who  identified   specific  diseases  and




symptoms  as  being likely   to   result   from  exposures  to  certain  air



pollutants.   Although  they  were  not  acting as  paid consultants  to our




study,  their  assistance  was most  important  in the  construction  of the




hypotheses we tested below.








1.5  Summary and Plan of this Report



The  study described  in this report  was an  ambitious one.   It  used an



unusually large and detailed individual data set; combined these individual




data  with  air  pollution  and  meteorological  data  that  were  as  finely




resolved in time and space as the SAROAD network,  the NOAA system and the




HIS permitted; included explanatory variables which, to our knowledge, were




not  used before;  experimented  with  a number  of  functional  forms,  model




specifications,  and  definitions  or  measurements  of  individual  health




status; and reported the results  of a  great many sensitivity tests designed

-------
                                   1-21






to  shed  light on  the confidence  one can  place in  the results  reported



below.




     Nevertheless,   it is  important  to  keep  this   study  in  its  proper




persepctive.  Because it is  an  epidemiological  study,  it suffers  from some




of  the  problems  identified earlier.  For instance,  although   individuals




were matched  to  the air pollution  monitors  nearest their  homes,  we still




lacked information  on their daily  activity  patterns.    Thus, we  could not




characterize precisely their exposure to ambient pollutant  concentrations.




Even if  we  had,  potential exposure to  indoor  pollutants would  still have




complicated the interpretation of our results.  Until personal monitors are




a practical solution, however,  some imperfection in characterizing exposure



and  dosage must be  expected.   Thus,  our  results  must  be  interpreted




carefully in view of the  imperfect exposure measures  we used.




     Second, in  spite  of  the detaiLed socioeionomic data contained  in the




HIS, we  lacked data on certain individual characteristics  or  habits which




may play an important  role in determining acute  or  chronic health status.



For instance, individuals  surveyed  by NCHS were  asked  nothing about their




exercise or  sleeping  habits,  dietary practices,  or  alcohol  consumption,




even  though  all  of  these  may influence  health  in  one  way or  another.




Similarly,  while we  had  information on  individuals'  occupations and the




industries  in which they are  employed,  it would have  been useful to have




data on  exposures   to  hazardous  substances in  the workplace.   In some of



these instances,  we constructed crude proxies for these factors.   But these




proxies  were  second-best   alternatives   to  more  detailed  information.




Throughout  the study  we  point  to respects in which  the estimates  could be

-------
                                   1-22






improved by additional information.  Nevertheless, we must emphasize at the




outset that we lacked perfect control measures.




     Because of  these (almost inevitable)  limitations,  as well  as others




identified below, the results of this  study must  be carefully interpreted.




It  is  important when  reading  the  following   chapters   to  pay  as  much




attention  to  the  caveats  about   our   estimates   as   to  the  estimates




themselves.  Similarly, one must pay close  attention to  the sensitivity of




a  particular  result  to   the   form  of   the   estimating  equation,  the




specification  of   the   model,   the  measurement   of  the  dependent  and




independent  variables,  and  the   sample  over   which   the  results  were




estimated.




     No single epideiniological or other kind of  study can ever come up with




the  "right"  answer.   Thus, the  results  of  this study  must  be viewed




alongside  those  from  other epidemiological,  clinical,   and  toxicological




studies.   One  can  have more confidence in  results  emerging from a variety




of  different  kinds  of  studies   than in  results  specific to  a particular




study or class of studies.




     The plan  of this report is as  follows.  Chapter  2  discusses the data




used in this report, and is divided into five sections.  The first of these




deals with the Health Interview survey, and goes into some detail about the




main survey (in  particular the definitions of acute and chronic morbidity),




the  smoking supplement,   and the  supplement  on  residential  mobility.   A




second  section  deals  with the  air pollution  data used  in  the analysis.




Included  is  a  discussion  of the  EPA  air  pollution monitoring  data,  the




completeness and quality assurance methods  to  used to screen  these data,

-------
                                   1-23






the procedure we used to create  a  measure  of pollution exposure over time,




our approach  to  naming the  pollution  variables in the  regressions  below,




and the procedure used to match individuals in the HIS to the air pollution




monitoring sites.  Two sections then follow on the meteorological and other




data employed in our analysis  and  the  purposes  for which they are used.  A




final section takes up the  question  of sample selection bias in the method




used to create our final sample of adults and children.




     Chapter 3 of the report  presents  the  methodology used in our analysis



of acute  and chronic morbidity.   Included there is:    a  discussion  of the




dependent  variables  used in  the analysis and  the  way in which  they were




created  from  the  HIS   data  tapes;   a  presentation  of  the  independent



variables  used  to  help  explain  acute  and  chronic  morbidity   (and  a




discussion about  possible  collinearity between them); and a  review  of the




estimating techniques employed in the analysis.




     In  Chapter  U  are  presented  the  results of  our  study.    Separate




sections are devoted to  the  analysis  of  acute morbidity due to all causes,




acute   morbidity  due   specifically  to   respiratory   disease,   chronic



respiratory  disease,  and   cardiovascular  and  other  diseases.    We  also




discuss in Chapter 4 the diagnostic tests we have conducted to identify any



collinearity that may exist between independent variables.




     Finally, in Chapter 5 we present some estimates of the changes in both




acute and  chronic health status that might  result  from changes  in ambient




ozone  concentrations  in the  urban  areas  of  the U.S.   These  predicted




changes might form the basis for estimates—in  dollar terms, perhaps—of

-------
                                   1-24






the benefits (or health costs) that might  result  from alternative National




Ambient Air Quality Standards for ozone.

-------
                                   1-25
                                 Footnotes





      Technically,  the executive  order  applies to the regulatory  agencies


governed  by  4U  United  States  Code  3502(1),  excluding  those  agencies


specified in 4U United States Code 3502(10).


      See Schneidennan, Mantel,  and Brown (1975).

     •3
      On sensitive  populations,  see Friedman  (1981).


      For  a  general   introduction to   epidemiology,  see  Lilienfeld  and


Lilienfeld (1980).


      For example,  see Lave and Seskin (1979).


     "See Whittemore and  Korn (1980),  for an  example.

     7
     'See Seskin (1979),  for instance.


     "For an introduction to and discussion of  indoor  air pollution and-its


possible effects on health, see National Academy of Sciences (1981) and the


General Accounting Office (1980).

-------
                                   1-26
                                References
Friedman,  Robert,   Sensitive   Populations  and  Environmental   Standards
     (Washington,  D.C.:   The  Conservation Foundation),  1981.

General  Accounting  Office,  "Indoor  Air  Pollution:  An  Emerging Health
     Problem," report no. CED-80-111  (1980).

Lave, Lester and Eugene Seskin, Air Pollution and Human  Health  (Baltimore,
     Md.:  The Johns Hopkins  University  Press),  1977.

Lilienfeld, Abraham  and David Lilienfeld, Foundations of Epidemiology  (New
     York:  Oxford University Press),  1980.

National  Academy   of   Sciences,   Indoor  Pollutants   (Washington,  D.C.:
     National Academy Press)  1981.

Schneiderman,  Marvin,  Nathan Mantel,  and  Charles  Brown,  "From  Mouse  to
     Man—or  How   to Get  from  the  Laboratory  to  Park Avenue  and  59th
     Street," Annals of  the  New York  Academy of Sciences,  vol. 246  (1975)
     pp. 237-248.

Seskin,  Eugene,  "An  Analysis  of  Some  Short-Term  Health  Effects  of  Air
     Pollution in the Washington, D.C. Metropolitan  Area,"  Journal of Urban
     Economics, vol. 6 (1979) pp. 275-91.

Whittemore,  Alice  and Edward Korn, "Asthma and Air  Pollution  in the  Los
     Angeles Area," American Journal of Public Health, vol.  70  (1980),  pp.
     687-96.

-------
                               Chapter 2
                                 DATA
2.1  Overview




The purpose of the data  used  in  this  study is to assist in determining




the  relationships,  if   any,  between  individuals'  acute  and  chronic




morbidity and  their  exposure to air  pollutants,  while controlling for




other factors.   As  Chapter  1 pointed  out, this  is not  an  easy task



given imperfections in the data,  problems in measuring exposure, and an




array  of  other  potentially  confounding  factors.    Nonetheless,  we.




believe the  data used in  this  study improve upon those  used in prior




epidemiologic investigations in several important respects.




      The basic  approach  in  this study is  the  following.   A survey of



individuals' health  and  socioeconomic status serves  as  the foundation




of  the  data base.   With  the  help of  the National  Center  for Health




Statistics  we  matched each  of  those interviewed in the  1979 Health




Interview Survey (HIS)  to the nearest  ten monitors  for  each of eight




different  pollutants by   assuming  that  each   household  lived  at  the




centroid  of the census  tract  in  which  its   house  or apartment  was




located.   This  was  accomplished  via a  computer algorithm  that  uses




input  data  on  the  geographic coordinates of  census tracts  from  the




Urban Atlas  Files of  the Bureau  of Census,  as  well  as  data  on  the

-------
                                  2-2
geographic  coordinates  of  air  pollution  monitors  from  the  EPA's




Directory of Air  Quality  Monitoring Sites.   Details  of  the procedure




are described later in this chapter.




      Because of the need  to preserve  confidentiality,  we could not be




given either the  address  or census  tracts of individuals  in  the HIS.



To  get  around  this  problem we  provided  the  NCHS  with  a  tape  which




matched each census  tract  in every SMSA in  the United States  with the




nearest ten monitors for each of eight different pollutants.  Using the




census tract identifiers on  their  "master" HIS  data tape, NCHS matched




individuals to monitors and included this monitoring information on the




HIS tape  they  sent us  (a  tape  which did not include  the census  tract




identifiers').   Matching weather stations  and  other area-specific data




was accomplished at the same time.




      Finally,   pollution  data   gathered  from  these  stations   were



summarized  and  combined with the  individual health and  socioeconomic




data.   The pollution  data were  averaged  over  three  periods:    the




two-week  period  in  1979  for  which  data  were  available  on  each




individual's acute morbidity; the entire year 1979; and, in some cases,




the  period  197^-79  (the  uses  of  these  data are explained  below).   A



similar procedure  was  used for  matching summarized meteorological data




to the individuals interviewed by NCHS.




      Two   distinct   groups  of   individuals  were  analyzed  in  this



study—men  and  women aged  seventeen and  above  (designated "adults"),




and children and  adolescents under  seventeen  (designated "children").




Both groups were drawn from the 1979 Health Interview Survey.  This was

-------
                                  2-3






a survey of  110,530  individuals  of all ages  and  from all geographical




regions of  the  United States.   In forming the "adults"  data  set,  the




original sample of  110,530  was  first cut by  requiring that all adults




had been given the supplemental  smoking survey, since estimation of the




effects of  air  pollution on respiratory and  cardiovascular health was




the  primary  focus   of   the  study.    The   smoking  supplement  was




administered  to  approximately  one-third  of   the  adults  in  the  main




survey, or 26,271  individuals.  This number was reduced further because




air  pollution  data  were  lacking  for  a  portion  of  these  adults




(described  below).    This  reduced  the  number  of  adults   to 14,416.




Finally,  five  individuals  with  internally  inconsistent  data  were



deleted, bringing the number  of  adults  to  be  studied to  14,4^1.  Table




2-1 indicates the distribution across U.S.  metropolitan areas  of those




in the final RFF sample of adults.




      The data set for "children" was created the same way, although no




observations  were lost  for  lack  of smoking data  since  the smoking




supplement was only administered to those aged seventeen or above.  The



number of children used in the analysis below is 15,711.  Later in this




chapter we discuss the possible bias introduced by our sample  selection




procedures for adults and children.

-------
2-4
Table 2-1.
SMSA
Number
0080
0160
0200
0240
0320
0360
0400
0460
0480
0520
0600
0640
0680
C720
0760.
0800
0840
0920
0960
1000
1120
1160
1170
1200
1240
1280
1320
1440
1520
1600
1640
1680
1720
1740
1760
1800
1840
1880
1920
1960
2000
2080
Residence of Individuals in RFF's
Location of SMSA
Akron, Ohio
Albany-Schenectady-Troy, NY
Albuquerque, NM
Allentown-Bethlehem-Easton. PA
Amarillo, TX
Anaheim-Santa Ana-Garden Grove ,
Anderson, IN
Appleton-Oshkosh, WI
Ashville, NC
Atlanta, GA
Augusta, GA-SC
Austin, TX
Bakersfield, CA
Baltimore, MD
Baton Rouge , LA
Bay City, MI
Beaumont-Port Arthur-Orange , TX
Biloxi-Gulf port , MS
Binghamton, NY-PA
Birmingham, AL
Boston, MA
Bridgeport, CT
Bristol , CT
Brockton, MA
Brownsville-Harlingen-San Benito
Buffalo, NY
Canton, OH
Charleston, SC
Charlotte, NC
Chicago, IL
Cincinnati, OH-IN-KY
Cleveland, OK
Colorado Springs , CO
Columbia, MO
Columbia, SC
Columbus , GA-AL
Columbus , OH
Corpus Christi , TX
Dallas-Ft. Worth, TX
Adults Data Set
Number of
Individuals
65
82
31
69
38
CA 170
65
30
64
146
26
37
40
199
34
34
30
50
38
93
276
37
c
25
, TX 36
145
42
21
47
703
157
218
49
45
32
19
92
42
193
Davenport-Rock Island-Moline, IA-IL 40
Dayton, OH
Denver, CO
89
124

Percent
of Total
0.45
0.57
0.22
0.48
0.26
1.18
0.45
0.21
0.44
1.01
0.18
0.26
0.28
1.38
0.24
0.24
0.21
0.35
0.26
0.64
1.91
0.26
0.04
0.17
0.25
1.00
0.29
0.15
0.33
4.87
1.09
1.51
0.34
0.31
0.22
0.13
0.64
0.29
1.34
0.28
0.62
0.86

-------
2-5
SMSA
Number
2120
2160
2240
2320
2360
2400
2440
2480
2520
2600
2640
2680
2760
2840
2880
2920
2960
3000
3040
3080
3120
3160
3240
3230
3320
3360
3400
3440
3480
3520
3560
3600
3640
3680
3720
3760
3840
3880
4000
4040
4120
4160
4280
4320
Location of SMSA Number of
Individuals
Des Moines , IA
Detroit, MI
Duluth-Superior, MN-WI
El Paso, TX
Erie, PA
Eugene, OR
Evans ville, IN-KY
Fall River, MA-RI
Fargo-Moorhead , MN-ND
Fitchburg-Leominster , MA
Flint, MI
Ft. Lauderdale-Hollywood, FL
Ft. Wayne, IN
Fresno, CA
Gads den, AZ
Galveston-Texas City, TX
Gary-Hammond-East Chicago, IN
Grand Rapids , MI
Great Falls, MT
Green Bay, WI
Greensboro-Winston Salem, NC
Greensville, SC
Harrisburg, PA
Hartford, CT
Honolulu, HI
Houston, TX
Huntington-Ashland, OH-WVA-KY
Huntsville, AL
Indianapolis, IN
Jackson, MI
Jackson, MS
Jacksonville, FL
Jersey City, NJ
Johnstown, PA
Kalamazoo, MI
Kansas City, KS-MO
Knoxville, TN
Lafayette, LA
Lancaster, PA
Lansing-East Lansing, MI
Las Vegas, NV
Lawrence-Haverhill , MA-HN
Lexington, KY
Lima, OH
26
407
16
49
31
29
22
22
48
14
57
73
20
51
42
40
51
53
34
52
66
45
34
70
69
230
32
33
103
37
24
58
61
24
39
132
39
45
44
46
35
30
41
35
Percent
of Total
0.18
2.82
0.11
0.34
0.22
0.20
0.15
0.15
0.33
0.10
0.40
0.51
0.14
0.35
0.29
0.28
0.35
0.37
0.24
0.36
0.46
0.31
0.24
0.49
0.48
1.59
0.22
0.23
0.71
0.26
0.17
0.40
0.42
0.17
0.27
0.91
0.27
0.31
0.31
0.32
0.24
0.21
0.28
0.24

-------
2-6
SMSA
Number
4400
4440
4480
4520
4560
4640
4720
4760
4800
4880
4920
4960
5000
5040
5080
5120
5160
5360
5400
5^40
5480
5560
5601-08
5640
5680
5720
5760
5880
5920
5960
6000
6040
6080
6120
6160
6200
6280
6320
6400
6440
6480
6600
6640
6680
Location of SMSA
Little Rock, N. Little Rock, AR
Lorain-Elyria, OH
Los Angeles-Long Beach, CA
Louisville, IN-KY
Lowell, MA
Lynchburg, VA
Madison, WI
Manchester, NH
Mansfield, OH
McAllen-Pharr-Edinburg, TX
Memphis, TN-AR
Meridian, CT
Miami, FL
Midland, TX
Milwaukee, WI
Minneapolis-St. Paul, MN
Mobile, AL
Nashville, TN
New Bedford, MA
New Britain, CT
New Haven-West Haven, CT
New Orleans , LA
New York, NY
Newark, NJ
Newport News-Hampton, VA
Norfolk-Portsmouth, VA-NC
Norwalk, CT
Oklahoma City, OK
Omaha, NE-IA
Orlando, FL
Oxnard-Ventura, CA
Patterson-Clifton-Passaic, NJ
Pensacola, FL
Peoria, IL
Philadelphia, PA-NJ
Phoenix, AZ
Pittsburgh, PA
Pittsfield, MA
Portland, ME
Portland, OR-WA
Providence-Pawtucket-Warwick, RI
Racine, WI
Raleigh -Durham, NC
Reading, PA
Number of
Individuals
38
27
754
88
1Q
38
36
13
46
37
76
4
142
41
151
180
38
68
17
11
39
116
1,184
210
30
64
9
72
43
41
38
178
31
33
483
88
279
11
47
70
98
39
16
32
Percent
of Total
0.26
0.19
5.22
0.61
0.31
0.26
0.25
0.09
0.32
0.26
0.53
0.03
0.98
0.28
1.05
1.25
0.26
0.47
0.12
0.08
0.27
0.80
8.20
1.45
0.21
0.44
0.06
0.50
0.30
0.28
0.26
1.23
0.22
0.23
3.35
0.61
1.93
0.08
0.33
0.49
0.68
0.27
0.11
0.22

-------
2-7
SMSA
Number
6760
6800
6840
6880
6920
7040
7120
7160
7240
7280
7320
7360
7400
7480
7500
752C
7600
7680
7800
7840
8000
S040
8080
8120
8160
8200
8280
8320
8360
8400
8440
8480
8520
8560
8680
8720
8840
8880
8920
8960
9000
9040
9120
9160
Location of SMSA Number of
Individuals
Richmond, VA
Roanoke , VA
Rochester, NY
Rockford, IL
Sacramento, CA
St. Louis, MO-IL
Salinas-Monterrey, CA
Salt Lake City, UT
San Antonio, TX
San Bernardino-Riverside-Ontario, CA
San Diego , CA
San Francisco-Oakland, CA
San Jose, CA
Santa Barbara, CA
Santa Rosa, CA
Savannah, GA
Seattle-Everett, WA
Shr eve port , LA
South Bend, IN
Spokane , WA
Springfield-Chicopee-Holyoke, MA-CT
Stamford, CT
Steubensville-Weirton, OH-WVA
Stockton, CA
Syracuse, NY
T a coma, WA
Tampa-St. Petersburg, FL
Terre Haute, IN
Texarkana, TX-AR
Toledo, OH-MI
Washington, DC-VA-MD
Trenton, NJ
Tucson, AZ
Tulsa, OK
Utica-Rome, NY
Vallejo-Napa, CA
Washington, DC-MD-VA
Water bury, CT
Waterloo , IA
West Palm Beach, FL
Wheeling, OH-WVA
Witchita, KS
Wilkes Barre-Hazelton, PA
Wilmington, DE-NJ-MD
27
26
104
30
82
217
27
52
96
134
146
328
61
43
48
44
170
29
28
29
57
32
14
35
70
37
90
50
38
67
36
28
39
51
31
21
258
26
37
37
13
35
38
51
Percent
of Total
0.19
0.18
0.72
0.21
0.57
1.50
0.19
0.36
0.67
0.93
1.0
2.27
0.42
0.30
- 0.33
0.31
1.18
0.20
0.19
0.20
0.40
0.22
0.10
0.24
0.49
0.26
0.62
0.35
0.26
0.46
0.25
0.19
0.27
0.35
0.22
0.15
1.79
0.18
0.26
0.26
0.09
0.24
0.26
0.35

-------
2-8
SMSA
Number
9200
9240
9280
9320
Location of SMSA

Wilmington, NC
Worcester, MA
York, PA
Youngstown-Warren , OH
Number of
Individuals
41
29
39
58
Percent
of Total
0.28
0.20
0.17
0.40

-------
                                  2-9


2.2  The Health Interview Survey

      2.2.1   Main Survey

      The 1979 Health Interview Survey was administered by the National

Center for Health Statistics (NCHS) of the Public Health Service of the

U.S. Department  of  Health and Human Services.   The  following general

description may be useful:
          The  National  Health  Interview  Survey utilizes  a
          questionnaire that obtains  information on personal
          and    demographic   characteristics,    illnesses,
          injuries,  impairments,   chronic  conditions,   and
          other health topics.   The population covered by the
          sample for the  National  Health Interview  Survey is
          the  civilian,   noninstitutionalized  population  of
          the  United  States  living  at  the  time  of  the
          interview.  The  sample does not include members of
          the  Armed  Forces  or  U.S.  nationals  living  in
          foreign countries.

          The   sampling   plan  of   the   survey  follows   a
          multistage  probability   design  which permits  a
          continuous     sampling     of     the     civilian
          noninstitutionalized   population  of   the   United
          States.  The sample is designed in such a way that
          the  sample of  households interviewed  each  week is
          representative  of the target  population  and  that
          weekly  samples  are   additive  over  time.     This
          feature  of  the  design  permits  both  continuous
          measurement of  characteristics  of samples and more
          detailed  analysis of  less  common characteristics
          and  smaller   categories   of health-related  items.
          The  continuous  collection  has  administrative  and
          operational advantages as  well as technical assets
          since  it  permits fieldwork to be handled  with an
          experienced,  stable staff.

          The overall sample was designed so that tabulations
          can  be   provided for   each  'of  the   four  major
          geographic  regions  and  for  selected  places  of
          residence in the United States.

          The  first  stage  of the  sample  design  consists  of
          drawing  a sample of  376  primary sampling  units
          (PSU's)  from  approximately  1,900  geographically

-------
                                 2-10
          defined PSU's.   A PSU consists  of a county,  a small
          group  of   contiguous   counties,  or   a   standard
          metropolitan   statistical   area.      The    PSU's
          collectively  cover   the   fifty  states  and  the
          District of Columbia.

          With   no   loss  in   general   understanding,   the
          remaining  stages  can  be  combined  and treated  in
          this  discussion as  the  ultimate stage.    Within
          PSU's, then, ultimate  stage units  called  segments
          are defined in  such a manner that each segment
          contains an expected six households.   Three  general
          types of segments are used.

               -  Area   segments   which   are   defined
                  geographically

               -  List   segments,   using   1970   census
                  registers as the frame

               -  Permit   segments,  using  updated  lists
                  of building  permits issued in  sample
                  PSU's since 1970

          Census address  listings were used  for  all  areas of
          the country  where  addresses -were  well  defined and
          could  be used  to locate housing units.   In  general .
          the list frame  included the larger urban areas of
          the United  States from  which  about two-thirds  of
          the NHIS sample was selected.

          The  usual  NHIS  sample  consists  of  approximately
          12,000  segments  containing  about 50,000  assigned
          households, of  which 9»000 Were vacant, demolished,
          or  occupied  by  persons  not  in  the  scope   of  the
          survey.   The  U2,000  eligible   occupied  households
          yield  a   probability  sample   of  about   111,000
          persons.

      In 1979, there was  a total noninterview rate of  approximately 3.9

percent, consisting of individuals refusing to respond or who could not

be located at home after  repeated calls.

-------
                                 2-11






      This is a  cursory  description of the HIS;  the  interested reader




is referred to more detailed information available elsewhere.




      The  HIS   provides  a   rich   description  of  the  socioecononiic




characteristics  of  those surveyed.   Each  year information  is  sought




about   individuals'   health   status   and   household   and   family




characteristics, educational  attainment,  income,  occupation, number of




hours  worked,  and  many  other  subjects  of  potential  interest  to




researchers.  Furthermore, the 1979 HIS contained detailed supplemental




surveys  on smoking  behavior and  residential  history,  each  going to




one-third  of  the adults  surveyed  (the selection of subsamples produced




considerable  overlap between  those  who  received the smoking survey and




those who  received  the residential  migration survey.)   Appendix  A to




this report presents the  1979 Health Interview Survey in its entirety,




as well as the smoking and residential history supplements.




      It   is   for   the   provision   of   detailed  information   about




individuals'  health  status,  however, that the HIS  was  selected  as the




foundation  of our  data  base.  It  is designed  to provide  a detailed




picture  of interviewees'  health status  at  the time  of  the interview,




rather than broader and more general information.  Because the data are




self reported, response biases are  possible with respect to health and




other  types  of  information  that  could  confound  estimation results.




Further, because the HIS was not  designed solely  for  the purposes of




epidemiologic research,  but  rather to  provide  a  detailed  picture of




health in  the U.S., it  does  not  elicit  certain  information one would




collect  in a  survey designed explicitly  for  epidemiological purposes.

-------
                                 2-12






Because  of  the  size  of  the  HIS  sample,  however,  and  the  details




available on individuals' health and other  characteristics,  the HIS is




a very rich data base.




      As discussed above, it is difficult  to estimate the  determinants




of acute and chronic morbidity.  In  fact  it is hard to define sickness




or  health,  much less  summarize it  by a  quantitative or  qualitative




index that can be utilized in  statistical  estimation.5   The  problem is




compounded when using the HIS because health measures are self-reported




rather  than  physician-diagnosed.   We  now  turn to  a  discussion of the




health  status  measurements  available  in  the HIS  and utilized  in the




present study.




      It is  not  uncommon in  health surveys  to  ask individuals if they




consider their health to be excellent,  good, fair, or poor.  The "EGFP"




question,  as  it has come  to  be known,  is  included  in the HIS.   The




response to  such a  question  conveys too little  objective information




about  health,  however,  and  two  individuals  in  equivalent   states  of




health might answer differently for totally unrelated reasons.  Indeed,




the same individual  might answer the question differently—without any




change  in  objective health  status—at  two different times  during the




same day.°




      The  HIS,   however,   goes  much  further  in  eliciting  health




information.    Survey  questions  are  aimed at  ascertaining  not  only



whether  the  respondent did or did  not have  any  of a  detailed set of




chronic or acute conditions  or impairments  defined below,  but also the




severity of  such conditions.   Severity  is quantified  in  a  variety of

-------
                                 2-13


ways,  including  time  since  onset of  condition,  duration  of  activity

limitation, and degree of physical disability  due  to limitations.   The

following  describes  the  measures of  health  status  compiled  in  that

survey.



      Terms Relating _t£ Conditions

           Condition—A morbidity condition, or simply a condition, is
      any entry on the questionnaire that  describes  a departure from a
      state  of physical  or  mental  well-being.    It  results  from  a
      positive  response   to   one  of  a  series of  "medical-disability
      impact"  or  "illness-recall"  questions.    In  the  coding  and
      tabulating   process,   conditions  are   selected  or  classified
      according to a number of different criteria (such as whether they
      were medically attended, whether they resulted in disability, or
      whether  they were  acute or chronic) or  according to the  type of
      disease, injury, impairment, or symptom reported.

           Conditions  except  impairments  are   classified   by  type
      according   to   the   ninth   revision  of   the   International
      Classification of Diseases, with certain modifications adopted to
      make the code more suitable for a household interview survey.

           Chronic Condition—A  condition  is considered chronic if (1)
      the condition is described by the respondent as having been first
      noticed more than three months before the week of the interview,
      or (2) it is one of the following conditions always classified as
      chronic regardless  of the onset.

                         Tuberculosis
                         Neoplasms (benign and malignant)
                         Diseases of the thyroid gland
                         Diabetes
                         Gout
                         Psychoses and certain other  diseases  of
                            the central nervous system
                         Multiple  sclerosis   and  certain  other
                            diseases of the central nervous system
                         Certain diseases of the eye
                         Certain   diseases   of   the   circulatory
                            system   (includes   rheumatic   fever,
                            hypertension,  stroke,  and  all  heart
                            conditions)
                         Emphysema,   asthma,   hay   fever,   and
                            bronchiectasis

-------
                           2-m
                   Ulcers and certain  other  diseases  of the
                      esophagus, stomach, and duodenum
                   Hernia  of   abdominal  cavity  (includes
                      rupture)
                   Gastroenteritis    and    colitis     (with
                      exceptions)
                   Calculus  of   kidney,   ureter,  and  other
                      parts of the urinary system
                   Diseases of the prostate
                   Chronic cystic diseases of the breast
                   Eczema and certain other dermatitis
                   Arthritis and rheumatism
                   Cyst of the bone (except jaw)
                   All congenital anomalies

     Impairments—Impairments are  chronic or  permanent defects,
usually static  in nature, that  result from disease,  injury, or
congenital  malformation.   They  represent  decrease  or loss of
ability to  perform  various functions, particularly those  of the
musculoskeletal system and the sense organs.

     Onset  of_ condition—A condition  is considered to have had
its onset when it was  first  noticed.   This  could be the time the
person first felt sick or became injured, or it could be the  time
when the person or family was first  told by a physician that the
person had  a condition  of which  he  or  she had  been  previously
unaware.
Terms_ Relating to Disability

     Disability—Disability is  the  general  term used to describe
any temporary or long-term reduction  of  a person's activity as a
result of an acute or chronic condition.

     Disability  day—Short-term disability  days  are  classified
according to  whether they are  days of  restricted activity, bed
days, hospital  days, work-loss  days,  or school-loss  days.   All
hospital  days  are,  by  definition, days  of bed  disability; all
days  of  bed  disability  are, by definition, days  of  restricted
activity.  The  converse  form of these statements  is,  of course,
not  true.   Days  lost  from  work and  days  lost from  school are
special   terms   that  apply  to  the   working  and  school-age
populations only  but these too  are days  of restricted activity.
Hence, "days  of restricted activity"   is  the most  inclusive term
used to describe disability days.

     Restricted-activity day—A  day of restricted activity is one
on which  a  person  cuts  down on his or  her  usual  activities for
the whole of  that  day because  of  an  illness or an injury.   The

-------
                           2-15
term  "usual  activities"  for any  day  means  the  things  that  a
person  would ordinarily  do  on  that  day.    For  children under
school age, usual activities depend on whatever the usual  pattern
is for the child's day, which will in turn be affected by  the age
of the  child, weather conditions,  and  so forth.   For retired or
elderly  persons,  usual  activities  might  consist  of almost  no
activity, but cutting  down  on even a small  amount for as much as
a  day  would constitute  restricted  activity.    On Sundays  or
holidays, usual activities are the things the person usually does
on such  days—going to church, playing golf,  visiting friends or
relatives, or staying home and listening to  the radio,  reading,
looking   at   television,   and  so  forth.     Persons  who  have
permanently  reduced  their usual  activities because  of  a  chronic
condition might not  report  any restricted-activity days  during a
two-week  period.   Therefore, absence of restricted-activity days
does not  imply normal  health.

     Restricted activity  does not  imply complete inactivity, but
it does  imply only the minimum  of usual activities.  A  special
nap for  an  hour after lunch  does  not  constitute cutting  down on
usual activities, nor  does  the elimination  of a heavy chore such
as cleaning  ashes  out of  the furnace  or hanging the wash.  If a
farmer  or housewife  carries on  only  the  minimum of the day's
chores, however, this  is a  day of  restricted activity.

     A day spent in bed or  a day home from work or school  because
of illness or injury is, of course, a restricted-activity  day.

     Bed-disability  day—A  day  of  (bed)-disability  is  one  on
which a person stays in bed for all or most of the day because of
a specific illness or  injury.  All or most  of the day is  defined
as more  than  half  of  the daylight  hours.   All hospital  days for
inpatients are  considered to be  days  of bed  disability even if
the patient was not actually in bed at  the hospital.

     Work-loss  day—A day  lost  from  work  is  a day on which a
person  did not  work at his  job or business  for at least  half of
his normal workday because  of a  specific illness or injury.  The
number  of days lost   from  work  is  determined only  for  persons
seventeen years  of age and over  who  reported  that  at  any time
during  the  two-week period covered  by  the  interview they either
worked at or had a job or business.

     School-loss day—A  day lost  from  school  is a normal school
day on which a  child  did  not attend school  because of a specific
illness  or   injury.    The  number  of  days  lost  from school  is
determined only for children six through sixteen years of  age.

-------
                                 2-16






      2.2.2  Smoking Supplement



      In its administration of the  annual Health Interview Survey, the




NCHS includes what  are termed rotating supplemental  surveys  which are




administered at  less  than  an annual  frequency.   One such  survey of




relevance  to the  present  study is  the  smoking supplemental  survey




included  as a  part  of  the   1979  HIS.   Due  to  the nature  of  our




study—where analysis  of  the effects  of  pollutants on  individuals'




health  must  necessarily   be  concentrated  largely  on   respiratory




conditions—the availability  of   detailed  information on  individuals'




smoking histories  is a  necessity if the  effects of  air  pollution on




health are to be properly assessed.




      Detailed  data   are   available  from   the   smoking   supplement.




Individuals' responses  to  questions about the  following were recorded




in the 1979.survey:




      1.  Smoking status (Never,  Former, Occasional, Current)




      2.  Current smoking status   (Occasional Smokers in 1)



      3.  Number of cigarettes now smoked a day




      4.  Age started smoking regularly




      5.  Number of cigarettes smoked a day at peak period




      6.  Last smoked regularly




      7.  Interval since last smoked regularly  (Former smokers in  1)




      8.  Number of brands smoked




      9.  Type of filter




     10.  Kind of-cigarette smoked  (plain, menthol)




     11.  Package type

-------
                                 2-17


     12.  Cigarette size

     13.  Ever seriously attempted to quit

     14.  Total number of serious attempts to quit

     15.  Number of serious attempts in past twelve months

     16.  Starting time of last attempt

     17.  Length of time stayed off cigarettes

     18.  Tar level of brand smoked most

     19.  Nicotine level of brand smoked most

      In  1979,  26,271  adults (aged seventeen and  above)  were surveyed

about their  smoking habits, this  representing  approximately one-third

of all  adults  interviewed  in the main survey.   A facsimile of the 1979

smoking  survey is  included  with the  facsimile  of the main  survey in

Appendix A to this report.
         *
      One of the  drawbacks of  the  smoking supplement  is that typically

only one  adult  in each household  (or  two  adults  if there were four or

more  adults   in  residence)  was  surveyed  about  his   or  her  smoking

behavior.    This  makes  it  difficult  to  estimate  the  effects  of

sidestream smoking on the health of children and other resident adults.

In assembling the data used in the present study, with the exception of

the  data on  children,  we  have  deleted  those individuals for  whom no

smoking  data were  collected.   In  the estimation  phase  of  the study,

this presents some problems.  For example, if a respondent reports him-

or herself  to  be  a nonsmoker,  there  exists the  possibility  that the

individual is exposed to sidestream smoke in the house, apart from any

-------
                                 2-18
sidestream  smoke  experienced  while out  of  the  house.   Trie  Surgeon

General of the United States has reported that:

          Children of smoking parents had more bronchitis and
          pneumonia during  the  first year  of  life;  and acute
          respiratory disease  accounted for  a  higher number
          of  restricted  activity  days  (1.1  days)   and  bed
          disability  days   (0.8  day)  in   children  whose
          families  smoked than  in  those whose families  did
          not.    A  reduction  in  exercise  tolerance  with
          exposure  to  sidestream  cigarette  smoke  has  been
          demonstrated in patients  with angina  pectoris, and
          a  decrease  in  small  airway  function  of the  lung
          equivalent  to  that  observed  in  light  smokers (one
          to  ten   cigarettes  a day)  has  been  reported  in
          adults who  never  smoked themselves nor  lived with
          smokers, but  who  were exposed to cigarette smoking
          in the workplace.0

      The results  of  the smoking data  should be  interpreted carefully

since  there is evidence of an increased underreporting  of cigarette

consumption over time.  One researcher has concluded:


          Data  from four major surveys, spanning the  years
          since  the  Surgeon  General's  Report,  suggest  a
          significant   reduction  in   rates   of   cigarette
          smoking.    These  data,   however,   conflict  with
          production  and  sales  data   which  record  only  a
          slight   reduction.       Explanations    for   this
          discrepancy  range from  problems  in  the  surveys'
          methodology    to    increased   underreporting   of
          cigarette  consumption because of  both a  growing
          awareness of  the threat  to  health and  the social
          undesirability of smoking.'

      Smoking  has  become  more  stigmatized, which  may  account  for

whatever response bias  exists.   Future research  should control for the

possible   errors-in-variables    problem  with   the    quantitative   or

qualitative smoking measures used in analysis.

-------
                                 2-19






      Table  2-2  summarizes  the   smoking habits  of  all  individuals




receiving   the    1979   HIS   smoking   supplement   as   well   as   the




characteristics  of  the  adults used  in  the present study.  The  latter




are members of  the  former  for whom  air  pollution  data were available.




It  is  clear  that  the  smoking  habits  of  our subsample  are  virtually




identical to those of the universe of individuals answering the smoking




supplement.








      2.2.3  Residential Mobility Supplement




      As  part of  the 1979  Health Interview Survey, a  subset  of  adults




was surveyed  about  their residential histories.   They were asked about




how long they had  been  at their present  address, how many times they




had moved in the past  three years, how  far  they had  moved,  and when




they had  moved.   Data were  summarized  for 25,519  individuals, most of




whom were also given the smoking supplemental survey.




      The  reason  for our  interest  in  information about individuals'




residential   histories   centers   on  the  analysis   of  the   possible




relationship  between air  pollution and  chronic morbidity.   Previous




investigations  of   this  relationship   have   proceeded  by  examining




associations  between   chronic   disease   and   current  air  pollution




concentrations  in the area where respondents  currently live.   Current




air pollution levels are taken as representative of lifetime exposures.

-------
                           2-20
Table 2.2.  Comparison of Smoking Habits of the General
            Population (as Sampled in the HIS) and Those in the
            HFF Subsample
                  Main Survey             RFF Subsample
                  (n = 26,271)             (n = 14,441)
Smoking Status

    % Never           i»3.1                     41.9
    » Occasional       2.2                      2.1
    % Former          18.0                     18.3
    % Current         30.5                     31.6
    % Unknown          6.2                      6.2

Number of Cigarettes
Per Pay, Present
Smokers, Median        20                       20

Age Started Smoking
Regularly, Median      17                       17

Tar Level of_ Brand
Smoked Most (mgs),
Median                16.9                     16.9

Nicotine Level of_
Brand Smoked Most
(mgs), Median          1.05                     1.05

-------
                                 2-21






But on  average individuals move  every five years  and thus live  in  a




nunber of  places  in  the  course of their  lifetimes.   Thus, a  chronic




illness  that  may  have  been  initiated  and/or  promoted  by  poor  air




quality in one area might be falsely associated with air quality in the




area where the individual is living at  the time of the study.   Data on




residential histories can be helpful  in avoiding this potential source




of spurious correlation.




      In  some of  the  analysis  below, we used  information  from  the




residential migration  survey to identify  individuals  who  had lived in




the  same  area  over  the   course  of   the exposure   periods  we  were




interested  in,  typically  five  to  six  years.    Using  summarized  air




pollution  data from  these  periods,   we   tested  hypotheses about  the




health status of individuals living in geographical proximity  to these




levels—rather than to pollution in some other geographical area.






      2.2.U  Documentation and  Cross Referencing of HIS Data




      Appendix B to this report  contains a listing of the variables




used in our analysis of acute and chronic morbidity that originate in




the HIS.  This list cross-references the variable of interest,  its




location on the HIS Public Use  Data Tape (available from the National




Center for Health Statistics),  and the corresponding question number on




the HIS itself.

-------
                                 2-22
2.3  Air Pollution Data

      2.3.1  SAROAD System

      The  pollution data  used  in the  present  study  came  from  two

sources.   The  first was the Environmental Protection  Agency's  Storage

and Retrieval of Aerometric Data (SAROAD) system from which we obtained

every hourly or 24-hour reading from every air pollution monitor in the

U.S. for the year  1979.  The  second source  was the National Aerometric

Data  Branch (NADB)  time  series  data,  also  maintained  by  EPA.   The

following is a description of the SAROAD system as of 1979:
          In  accordance  with requirements  of the  Clean  Air
          Act and U.S. Environmental  Protection Agency (EPA)
          regulations for State  Implementation  Plans (SIPs),
          ambient  air   quality  data  resulting   from  air
          monitoring operations  of  state,  local,  and Federal
          networks must be  reported each  calendar quarter to
          the  EPA.    The   EPA   Storage   and  Retrieval   of
          Aerometric Data (SAROAD)  format  is  the  established
          medium for transmittal  of air data  to EPA Regional
          Offices  within  45  days  after   the  end  of  each
          reporting  period.    EPA  Regional   Offices  must,
          within  an  additional  30 days,  forward  data they
          have received  to  the EPA Aeormetric  and Emissions
          Report System  (AEROS), of which  the  SAROAD system
          is  an  operational party.   AEROS  is managed by  the
          National  Air  Data Branch  (NADB),  Monitoring  and
          Data  Analysis  Division,  Office  of  Air  Quality
          Planning and Standards.1^

-------
                                 2-23


      In 1979, the SAROAD files contained  data  on  approximately 12,000

monitoring sites,  4,000 of  which were  then operational.   The  codes

identifying each site are uniquely defined by five  criteria:


             (i)   State Code
            (ii)   Area Code (city, or if not in city, county)
           (111)   Site Code (001-999)
            (iv)   Agency Code
                   A-E ... EPA Headquarters Groups
                   F   ... State Agency
                   G   ... County Agency
                   H   ... City Agency
                   I   ... District Agency
                   J   ... Private
                   K   ... Institution
                   L   ... Military
                   M   ... International Agency
                   N   ... Other Federal Nonmilitary Agency
                   P   ... EPA Regional Group
                   Q   ... World Meteorological Organization
                   R   ... World Health Organization
                   Z   ... Other
             (v)   Project Classification Code
                   01  ... Population-oriented
                   02  ... Source-oriented
                   03  ... Background
                   04  ... Complaint investigation
                   05  ... Special studies
                   06  ... Episode monitoring
                   08  ... Global surveillance
                   09  ... Duplicate sampling
                   10  ... Continuous Air Monitoring Program

Thus, for example, a site identifier of 310660004F01 indicates  that the

monitoring  site  is  located  in  New  Jersey  (31),  specifically  in

Burlington County  (0660), its  number  is  004,  it is operated by a state

agency  (F), and the monitor is population-oriented  (01).  An identifier

of  482890001F03  indicates  a location  in Virginia  (48),  specifically

Shenandoah National  Park  in Page  County  (2890), at  site  001,  operated

by a state agency (F), for background surveillance  (03).

-------
                                 2-2U






      Other considerations are the  type  of sampler or monitor  used in




the data collection procedure and the frequency with which the data are




collected  (or,  if  continuous,  are summarized).   Insofar  as  collection




methods are concerned, the number of different types of monitors varies




with the pollutant  in question.  In 1979 suspended particulate monitors




were all  of  the same type  (HI-VOL Gravimetric), for  example,  whereas




there were  twelve  types  of  sulfur dioxide  monitors,  eleven  types of




nitrogen dioxide monitors, and three kinds of carbon monoxide monitors.




Two  different  monitoring methods  were  used  for ozone:   instrumental




chemiluminescence  and  instrumental  ultraviolet,  with  approximately




twice as  many  monitoring sites  using  the former as the  latter.   Both




methods produce  data  at  one-hour frequencies, whereas for some of the




other pollutants, the data frequency varies with the monitoring process




being utilized.








      2.3.2  Data Integrity and Completeness



      Three  basic   statistics  were  used  in  our  study  to  summarize




individual pollution exposure over the two-week period for which we had




acute health  data  and  all  were  derived directly from the  SAROAD raw




data.   These  were the  average of the fourteen  highest  daily one-hour




readings  for  a  two-week reference  period;  the  highest  single hourly




reading over the reference period; and the average hourly concentration



over the  entire two-week period.  Two measures  of  the completeness of




the  data  at  each monitor were also  included  in the air pollution  data




we   matched  to   individuals.     The   first   was   the   number  of

-------
                                 2-25






days during  a  two-week reference period  for which there was  at  least




one hourly or  24-hour  reading (bounded between 0  and  14).   The second




was the  total  number of hourly  observations available from  a monitor




during a reference week (336 maximum).




      Tables  2-3 and 2-4 compare for two randomly chosen monitors  the




annual   data   derived   from  our   method   of  summarizing  pollutant




concentrations by two-week  reference periods.  Our summary data compare




exactly with the monitoring data summarized for each  site  in the NADB




"Quick Look"  Report (included following the monitoring data for each of




the two  sites).   The  annual summary  data  appear in  the  last  row of




Tables 2-3 and 2-4.



      In  addition to  our  concern   about   data  completeness, we  also




analyzed the  quality of  the data in an  effort  to identify miscoded or




otherwise inaccurate data.   Because ozone  is the  focus  of  this study,




the  ozone   data  were  subjected  to   the  most   detailed  analysis.




Nevertheless,  all  the  air  pollution data  were  screened using  one or




more outlier tests.  Air quality data  are usually screened by federal,




state, or local reporting   agencies  prior  to submittal of  the data to




the SAROAD  system.  Thus, it  is  not surprising  that the  ozone data




exhibited reasonable consistency.   Given more time the screening tests




applied to the ozone data could be applied  to any of the other

-------
                                          2-26
•e 3
c c
^ ^
VI <
C i
. -
~ —
a
«
£

~
i
^

;•
1

-

*,
J1
r

-^ *»
A j»
3 *
C.
IT ^
£ s




-
^
— •
-
c
^












&
Cj
d,
^
T t r- -o *x <%' r; «r rj — —
~ i .•*! r-j ri "^ "i «^! « « ^*i
i
u*
o ' c% M »r CN r: ?3 ry >. r^
o i CM r4 rj »*i rr ro f*5 r^ f*J
i
w i
— I ^ •<  r; -c n n o
Si — ^ — XO^O"^S =
^ICCCOOCOCC
£ :
1
1
»««cs«r««ein* —
«• r-i r; — — M —
i

IK-!rj9.-c
-• n rg -. — rj
:15iili';£
~ i
t •
•« rg •: ^ .-. -c o. c >


•? ^ rs — • o s it" r» *• o* ^ *^ ii
^COfjCSC^^ CK*«^^9hO^ Ow
CM «• M rj f4 P-< fM n n  — r>
*^"*^U")G~'**0I»O^G^ "J
rt'*^rjr4r^^'*i'*^^r*5r*i1^^r
o ^ o o v o cs o o o o o o


oooeeeooooees
o ^~ .- s c c r>» -c ;•> :: >.: c -c
C %fv rv ^ j*; *N a c r-* r; o^ ^ v.
ooeooooo — — o .— —
<»











rvo^-.csrj"*c— «c ^^"'C:
o*"O^. SO*-- rv*c< v^ ic x *r
oocoecoceceoc



oin-.esiftrjo-«c-!Ofvmo
*• r* *• r* t*w «• rg TX ^

— t

cc^^_^c.cc<:<: = =
r = cc«««zsi=-:-:
1 1 ; 1 ; ; ! , : 1 ! 1 !
u-5!"<3«-crg&-<":e'v«r — s
— - M — rj n «. r.1 r.
«<:c<£r^i.4.i«ce — r.

-------
                                                               2-27
             c
             fj
                                                                                     ")
                                                                                     •)
                                                                                                  .1
                                                                                                  -1
.t-
   111
u in
t. o
r o
o  -
      o
      o
                                             —     o
                                             •v     03
                                             (N     CM
                                                                                                  •o
                                                                                                  •c
                                                                                                  Cl
                                                                 O     o-
                                                                 —     a
                                                                                                                                  c
                                                                                                                                  [••;
 ;  OP
 IT  i.'
 1  -
•s.  -
                    >    •»
                    m    "5
                    OS
r*>    *r     *     v
o    o     o     o
                                                                                    r-i
                                                                                    a
                                                                                                               c

                                                                                                               o
c     o    c     e    o
in     L~    -o     <    T
s     —    rj     r-j    o
—     —    —     —    CN
O    O
s    in
*o    «n
                                                                               0
                                                                               e
                                                                 o
                                                                 c
                                                    o
                                                    0
co«ooc*oc     ~
o    o     u-     ®    ©    *     f>    in     e     c
O.O-c'^r^-COO'r^'
ooo'Oeo     —    —     o     o
a
~v
O
                                 SB
                                 ^
                                 O
                                       o     —     —    —
o        rv
—    —-co-
o-    -c     >c    •<.
o    c     o    o
                                                    in
                                                    ~a
                                                    o
                                                                              ri
                                                                              ••^
                                                                              
                                                           in
                                                                                                  <    ro
                                                                                                  —    ri
                                              5    E

                                              «    «
                                       c

                                       c
             «•     —    CD
                                 in
                                 (N
                                                                                     M



                                                                                     C
                                                                              LU




                                                                               I




                                                                              PJ
                                       i.    i.
                                       ^.    Ui
                                       tn    CR
                                                                                                                      u    u     u    u     c
                                                                                                  »    —     o    in     CM
                                                                                                  n                 —     CM
                          V.
                          r.

-------
                                                               2-28
c -. tn
u S 1 -O -C -0 -O
O 3 "? l *^ ^ "^ M
~~ ~ Si 1
1
1
j
I
*. !
8 Ut t CO C4 ^ &»
£• U"5 1 OJ O" f* fl
3 O 1 « *« — -.
00 1
«• * 1
1
1
i
1
t
t
1
1
1
31 ~ i
«M C 1
•. •*• 1
5 w 1
C *5 1 "-O f*) M -C
S *• 1 ^ *V FJ C»»
-> 1 — *« ~ O
z c i c c o c
ft S !
"" S i
W :
!

!
i
i #

5 i
u ;
C !
z ' : * o c e
C i If! V. O O
— i — 1 ^ V » -5
in T O O O O
— a
wee .
•«• '
• ,
J
I
— «t T i
3 m A i
s/ (y i
a _ £ i
=. tt Z i
e i
1
!
I
t
1
1
7 1
— e i
o r i « > n >
e. e • > r^ ^ o*

c " i o c b c
& s
z: >

i
t
i
i
IT fvi O- -O
! (M —
;
I
1 S> U U U
a, i e uJ uj uj
-> ' Z ,=, C. C.
i t
Ail t t l
l
r-i > -c PI
— «•• r-j


^ ^ ^ u
t = e e a.
t. s x s: s.
u & :
«j b
^» 3
1. ' -o rv 3 o.
C : ^- «• «r v
"O -c t T
fi -I ! c '

i i
i !
1 !
1 1
1 1
f-i t*l 1 » 1
a n < r*. i
" (N ' CC 1
1 -0 1
1 1
1 1
t I
1 1
! 1
. 1 1
i 1
I
' f
1 !
1 1
1 f
1 1
1 1
ffl — l O l
•• •« i rv i
— — 1 CM I
o o • o i
1 !
i 1
] 1
I ,
1 1
1
1
1 1
1 1
1 1
1 1
; l
I 1
0010!
^ ^ ! ^ I
**5 ••* t O i
O O i (N i
; i
, '
! i
1 !
! 1
J !
I i
I ;
, f
l i

1 ;

; i
1
; i
T :
i i
; 1
1 1
i t
i '•
f. CM i e i
O «r i rt •
IN M • '.'. '.
0 C : 0 i
i i
'. '
i l
l i
l t
l t
1 i
not i
ts n \ i
' !
1 !
U U 1 '
UJ -. 1 i
OH, 1
i u i
l lie;:
1 — !
O Px 1 -i 1
•• •• i ml
l «• i
! W I
U U I fC
UJ LL' i ->
s. e. i in ;
i re :
i ~ i
C i
b~ mi c i
                                                                                            in a

                                                                                            < u»
                                                                                             in
                                                                                             in in
                                                                                             — ui
                                                                                            O X
                                                                                            M <

                                                                                            5 ""



                                                                                            S "*

                                                                                            S Q
                                                                                            in *•
                                                                                            (M in
                                                                                            m in
                                                                                            -t <
                                                                                       —    > =
                                                                                            3 O
                                                                                            Z X
UJ
>•

O

a

UJ

                                                                                       u

                                                                                       *r
      3E
      Z N
                                                                              u


                                                                              u
                                                                              MJ
                                                                              e

                                                                              5
                                                                              w
-    a
                                                                                       S
                                                                                       u


                                                                                       a
                                                                                       IM
                                                                                       in

                                                                                       e
        5

        i
                                                                              in

                                                                              e

                                                                              &



                                                                              |

                                                                              o
                                                                                       u
                                                                                       in
                                                                                               O
                                                                                       u



                                                                                       O^^O^OF1^
              »»>O"»«>O>9'»
              h»r^b*r^r^r«r^*F«r^


              _ =       
-------
                                                                                    2-29
              c —   ;
                    til   i
                 <- O
                 r o
                 c  -
              S  re  it   i
                 c  n   i
              T _ tt   I
              T          1
              re -j  c.
             —  K  :   ;
                    c   i
l_  C
L;  >
>.  c
£ 3
                               ecor^-GCMri^-eifto^ino-o-mw-o-er^-co
                               CCOIfl»W!P)Bf>»!\MO™)»lrt>     o     —     r^cKOO     Ofr*r     «roo     ixfv»«     *;
                 •*  T   i      ooom-c-cinmrrvv^-ini-rv     r>.«;'                 !^^8C^«*CCtf3«*OmfVO»'Of5Or^.
   IT.   .                 !«P1           •«     ^     p<           —    —     P,i                  —     fN     CM           M     Pa     M

c «•   •                 !
w —   :                 :ZZ£a,ac.«a:C£fca:c:csiic:c>->-^>-
c;  '                 tiC<!o-iO-
Ul IT   i                 .                   —     PV    CM            ••     «.    Pg           __[••.



^      '         t       'cccc«i_uiujiiic^B:o-o~PJ"i«r'i:")e~pj

-------
                                                          2-30
 r ffl  I
3. a

~~ <£.
                 >    -»
                 PJ    ."I
                                  p.
                                  o
                                  n
                       PJ
                       n
                                                    in    —
                                                                                       .o    y,

                                                                                       n    "i
,s
   in
w in
- o
y e>
e  •
fs
PJ
rs
PJ
PJ
o
n
o-
PJ
                                                    rx
                                                    in
                                                    CN
-•    n
a    -o
CM    PI
..    -5
w    r»
PJ    PJ
3 « :
C - , — O 0» O PJQ
3 «* : r> -c eo n ^ o
-Ol'O O C O OS
S. Z i
s a <
? o ^ PJ «o PJ pi in «o « o» i*» » rs n © o ">>
* P* © Q3 ^ 53 ^* ^ ^ |/3 ^ rt p^ m ^|» i*^ ^ *^
SOSOOOOOOCOOOCOOOC




1
1
,
(
I

1
1
!
i
I
I
!
i

1
i


,
i

1
1

t
1


,


I

t
1



— z \ c c o o
•** * r* t\ ^o rv
*j "T i ^ o- e o
« « i/ C O,O w
'ri "" i
••* t
!
!
W
£ C '
— r "5 !
3 « * i
4i ft 1
ts ^ as i
r. «*  C
X £ l
1
I
1
i x «• — o
1 — PJ
! | | 2 S
« >
a i i i i i
]«•««»«
•• •« PJ
f ^j ^ ^ ^
i r r s r

w ^.
b. S i
a. a i
^ 3 '
Oi rj PJ pi p.
cococooooo«c sooccco
rs*.'«C--tnP2'*3«"'«~*C«'* — OCTf^rv'>*X"s-
ooco — •»NrMOr)Wo«o>rs-o5>oees
^ruinifi-o-o <«««««< in~5pji^pjp, PJ
ocooooooocoo oooooo o



«pj>mpjO'«cpj>-V>- »• l_ H> ^ t—
~~~)<<<«cocncacntn sscezzs
1 1 t 1 1 ] 1 1 1 | 1 1 1 | ! ] 1 | 1
PJ>-C")O C ^ O ^ P-< **? ^ li"5 «C rs- E ^ O •* PJ "• ^ t1^
p-ipiPjrjn^fOWMnrin prv^^r"^v

-------
                                   2-31
Ifi   *C   CN   3D
                 CM
                 n
C C "1 ' *.' f*, ™i rt n M ' PN •
z z si i
s ' ,
i !
,i : i
trt l ; i
am i o r\ -o *• -! o ' — '
1.0 i rv ca in o > < ' > i
3O i — IN M r, — -. . m :
e - : -o •
~ i i i
i . i
1 1 !
1 1 1
! 1 1
1 1
! '
si- i
—*••••: i
5-icB -• n tM o fi i M ,
Z-' i in •» CN o o ^ i r?,
c i — — — -« o e ' M i
CIO OO O O O • C
Oi C :i
X 0 it
U 1 :
1 1
' 1
1 !
a , i !
ST. 1 i
x z , e o c o c o • e .
«J T i u~ m *• » i"! r, mi
mm e o o o O Ol — .
TJ~ i 1 ;
-~ ! '
X ' 11
' 1 1
I l :
; i
^^ ' '
*J «d — • ,
3 rc rc i |
i> O l i
m -: Z l i!
ct| i"-----, -I
1 i ,
1 . ;
~ e ! I i
a * OdiT — »«•'>:;
— ir r< CM rj — — ; «•
CX OOOCOOiO
tt, Z : i
S ;
• t
i : i
i 1 !
l 1 1
i in c 4 >• ^ n o i '
'P; «• rj ri i i
! 1 '
• 1 1
r> u u u u u ' :
U i ^ ^ ^y j^ ^ m | j
-1 r B e a e a i
•» . la;
C- l i l f l t l g i
i rw o. -c n o rs i -j i
•" •» PJ — _• ' ul !
i — i
> — i
2 ^2 u U U • (t !
a< c — o LJ u; k* ' w i
^ '^SZAAAlg;^
a; A. i ^
>- * (B
i i -
5. . -0 rs S 0> O — . - •
c i «•«•»«• is 10 c

RIIOUAHIIIE D DYE-IS
U
u
in
c i
5 =
u
2 I
* 1 5
a ~j c
S£ = ^
u ac o a
K ^ — • tn
£ 0 M Q
S ~" ~ t-
-J U UJ —
- cr 
i i s
M B, p.
Z UJ
i £
O i
u
UJ
in
z
»*
r
u
n
UJ
1



>•
<
a
01
in
r
_/
a
in
A
m
>
i
5
_









r
a
V)
V
«
l/>
<
X
— 1
PN
in
«
•<
UJ
cr
§
WB
Ul
1
m
a


A1IUI03
LOCATIOII

O
Ul
V)

o
o

S
o

^
»y
Q
f^,
>i uiinouoti
<
^
9AMGAII3M CO
o
•J
M
1
e
0
o
o
CJ
( J
^

o
o

S
1
S
in
<
u>
S
a.
a
noi.-iioE co
1 MAItRlOO
e
0
a
o
f

o
f»

S
g
I
r^
tn
i
in
u
-i
u.
O
O
u
z
«.'
o
IZOOOZC9H J

o
e


8
2
•o
<
^
z
o
z
0
u
u
<
^
UJ
O
0
e
e
rj
^
«•

O
o

2
2
2
tn
ta-
i.
u
UJ
0
0
_
•M
O
f l'.8120007l

in o
r* O

2§
II
S 0
n n
u. •
O =
h. in
u
m —
c
* •
e
u
o <
U Q
0
U
O
. 13
O O
o o
o o
0 0
rj o
If,

«
.


S
S
r^
p»
*
in
2
B
MADISOII CU
ce
UJ
1
0
o
o
o
1

-------
                                 2-32






pollutant data series used in  this  study by simply modifying the basic




ozone quality control  statistics.   Based on  the  extensive analysis of




ozone and screening of  the other  data,  it appears that the air quality




data are not subject to serious irregularities.




      Ozone  exhibits  consistent   diurnal  and  seasonal  concentration




patterns depending  on  a wide  variety  of meteorological  factors.   The




tests selected to discover anomalies in  the  data  reflect this pattern




and may  be grouped  into  tests designed to  identify  implausibly high




concentrations  based  on  historical  data,  and  tests that  basically




evaluate the consistency of diurnal data.




      In this  study four  tests  were selected:   the  maximum  hour and




adjacent hour  tests,  and  two variants  of  the  spike test  based upon




percentage  and absolute differences in  adjacent  diurnal  hours.   Table




2-5 presents the critical values  used  in the four tests.   They reflect




recognized  seasonal and diurnal ozone  concentrations,  and check for an




upper limit for particular ozone  concentration.   It is well known that



ozone concentration varies  by both season  and  day  and  while  these




critical  values   are  somewhat subjective,  the   basic   motivation  in




applying   the   tests   is  to   identify  data  that   warrant  further




examination.




      The basic procedure for evaluation is relatively straightforward.




The maximum hourly  test simply searches for data  exceeding a maximum



upper bound.  The second test compares an upper limit on  the difference




between  adjacent  hourly ozone readings.   The  third  and fourth tests




compare the middle of three consecutive  hours in terms of absolute and

-------
                                     2-33
Table 2-5. Selected Quality Centre
Hourly Values (Concentr
Data
Stratification
Sunnier -Day
Summer -Night
Winter-Day
Winter-Night
Maximum Hour
Test
.501
.383
.255
• "53
1 Tests: Critical Values for 3AROAD
ation in Parts per Miixion)
Adjacent Hour
Test
.153
.102
.128
.102
SPIKE2
Test-1
.102
.051
.102
.051
0.
.5
SPIKE3
Test-2
300J
300S
300%
3002
'Reference weeks 15-44 are classified as ''summer," wniie 1-17 and 45-52 are
 "winter."

"Under this spike test, the second of three consecutive hourly readings is
 flagged if it exceeds either enapoint by tne stated value.

"'Under this spike test, the second of three consecutive hourly readings is
 flagged if it exceeds either endpoint Dy 3005.

-------
                                 2-34
percent change in hourly ozone concentrations.  Each of these tests are

fully described elsewhere, as are specific results applying these tests

to specific monitors.    The stratification of ozone data by season and

time of  day is given  in  column 1 of  Table 2-5, whereas  the  critical

values are reported in columns 2-5.

      There are  drawbacks to  these  tests.   For example, it  would be

preferable  if  the  critical  values  in Table  2-5  were available  for

separate  geographic   regions.     Furthermore,   the   critical  values

themselves  are  subjective.     Finally,   there   are   relatively  more

sophisticated  data  screening  tests  available.   However,  it  is  very

difficult  and  costly  to  obtain  and  evaluate  air  pollution  data  by

regional  grouping  let  alone  by  specific  monitor.    In our  analysis

critical  values were  only used  to  identify data that  warrant further
                                            *
study.   Finally,  more  sophisticated  tests are themselves  more costly

and  only  recently  are   being  applied   to  evaluate  SAROAD  data.

Obviously,  given  more  time  and  resources additional  tests  could  be

applied,  but  those  tests   conducted   give us   confidence  about  the

relative  consistency of the ozone data we used.

      Approximately  660  SAROAD  ozone  monitors   reporting   daily  data

during  1979 were screened using the four  tests  described  above.  This

includes  all  the  air   pollution   data   available  to  this  study—

approximately five million hourly observations for 1979.  The number of

ozone monitors  in  each state  that had  data  flagged by  the screening

tests is  presented  in Table 2-6.   We  checked each hourly observation

flagged by the screening tests.  In every case the reading was a

-------
                                   2-35
Table 2 5.  Number of Ozone Monitors Flagged in Data Screening Tests
S-ate
Maximum Hour
                                                  Test
Adjacent Cutlier
Arizona




Arkansas




California




Colorado




Connecticut




Delaware
             olumbia
Georgia
    ana




Kentucky




Maryland




Massachusetts




Missouri




Montana




Nebraska




Nevada




New Jersey

-------
                                    2-36



Table 2-6 ^continued)






                     	Test	




Stats                  Maximum Hour          Adjacent Outlier          Spike






New Mexico                                                               1




New York                                            1                   12




Ohio                                                                     3




Oklahoma                                                                 1




Pennsylvania                 1                       4                   13

-------
                                 >-37
plausible one and apparently not the result of a malfunctioning monitor




or  an error  in  coding  the  data.    The  specific  readings  that  were




flagged are available upon request.






      2.3-3  Multi-Year Averaged Data



      As discussed above,  several  different  measures  have been used to




summarize  and  verify the  raw air  pollution  data for  1979  from EPA's




SAROAD system.  This was done to provide a view of the overall exposure




to  ozone  and other  air  pollutants which  individuals received  in the




different  SMSAs around  the  United States.    We  have  also  summarized




individuals' exposures  to  air pollution over  a number  of  years using




long-term data from the air pollution monitors assigned to them.




      The summary data used in  this exercise are  those in the -National




Aerometric  Data Base, Yearly Frequency Distribution.   The  data  tape




made  available  to   us  included  annual  summary  measures  for  ozone,




suspended  particulates,  nitrogen  dioxide,  carbon  monoxide,  sulfur




dioxide, and lead,  for the years  1974 through 1980.   No sulfate  data




were  available.   For each pollutant,  the  method  of  collection, number



of annual observations, arithmetic and geometric means, and highest and




second  highest  annual  values were recorded.   It  was  also  reported




whether the  annual  summary measure reported  met  the EPA  criteria for




data  completeness.



      In attempting to compile multi-year summary measures of pollution




to  characterize  individuals'  long-term exposure to  ambient  levels, we




have  used only the data meeting EPA completeness criteria for the years

-------
                                 2-38






1974 through  1979.  We matched  on  a monitor-by-monitor basis such data




to  each  individual's  other recorded  data  in  a  method  completely




analogous  to the  matching  procedure  described  above.   It  will  be




recalled that the monitor matching is based on monitors in operation in




1979,  and  that  data  were  deleted  for  all monitors  more than  twenty



miles  from the  centroid of  the  census  tract in  which a  particular




individual  was   living.   For each  of  the  six  years  1974-1979,  the



summary  data were  averaged  over   all  monitors  within twenty  miles.




Finally, these six  annual averages were simply averaged  to  obtain the




summary  measures of  interest.   In each  of  the  averaging  processes,




missing  data  were ignored  and  averages were  calculated based  on the




existing  data.   We  recognize that  our method obscures the  variance




about  the  mean  at an individual monitor  over the course of one year,




between all the monitors within  twenty miles  for a given year,  and over




the  multi-year   "grand  averaging."   In  subsequent  work we  hope  to




include measures of dispersion as well as  central  tendency.




      An example is presented  to  clarify this procedure,   Table 2-7




contains   illustrative   annual   average  hourly   ozone  data  for  one




individual.   The first  row shows  the  distances  in  miles  to the ten




nearest ozone monitors for  the individual.  Rows  two  through seven are




the  1974 through 1979 concentrations at these monitors  for  data which




meet EPA completeness criteria.   Missing values are indicated  by dots.




The  final  column in  rows  two  through  seven  presents the  six  annual




averages,  subject to  the twenty-mile cutoff.   The figure  in the lower




right hand corner of the table shows the value derived as this

-------
                                     2-39


Table 2-7.  Summary of Long-term Annual Average Hourly Ozone Concentration for
            a Representative Individual


                      Annual Average Ozone Concentration


                                                                     Average of
Monitor                                                               monitors
no.      1     2     3     4     5     6     7     8     9     10      1-4l7

Distance  879  fTlTO  TO  2l792278  22792376  2876    3276

Year
1974      0.19   .......     .015    .       .019

1975       ........     .012

1976       .023  ......           .011           .023

197"       .020  ......     .018  .012    .       .020

"978       .019  .024  .013  ....     .013  .014    .022   .0187

1979       .     .     .014  .      .     .013   .029  .015  .       .       .014

                                                        Overall average:   .0189


      Only monitors 1-4 are used to compute the within-year and multi-year
averages in this example because monitors 5-10 are more than twenty miles from
the individual's home.  This is the procedure used in the creation of the
multi-year data.

-------
                                 2-UO






individual's  long-term annual  arithmetic  average  ozone  concentration




(.0189 in this case).




      The summary  measures  constructed from  the  NADB Yearly Frequency




Data are  as  follows.   For ozone, sulfur  dioxide, carbon monoxide, and




nitrogen  dioxide,  the six-year  averages  of both  the arithmetic means




and  the  second  yearly  high  values  were  computed.   For  suspended




particulates, the  average of  the geometric means  as well  as the second




yearly high  values were computed.   Also  calculated for each individual




were  the  number  of  monitor  values  used  in  calculating  each  yearly




average as well as the number of years used in calculating the overall




average.




      When contrasting the  multi-year annual  averages  with the annual




averages  calculated from  the  1979  data alone, one expects  the summary




to differ.   For example,  for  the 1U,U41  adults in our sample Table 2-8




shows  the simple  correlation matrix  of  the five  different  long-term




measures of ozone used in the present study.  The statistics calculated




from the  1979 hourly data  are  also subject  to  the twenty-mile cutoff




rule,  i.e.,  readings  from  monitors at a  distance of  twenty  miles or




greater  from the  individual's  census tract are  assigned a missing




value.  The ozone measures are indexed as follows:




      03(1) = Average daily 1 hr. maximum, 1979




      03(2) = Maximum hourly reading, 1979




      03(3) = Average hourly reading, 1979




      03(4) = Annual arithmetic average, 1974-79




      03(5) = Average of second-high yearly reading,  1974-79

-------
                                 2-41


03(1),  1 = 1,2,3,  are derived  from  the  raw  hourly data  for  1979 while

03(j), j = 4,5 are derived from the  NADB Yearly Frequency data using the

methods described in the  above paragraphs.   The  correlation matrix is

as follows:
      Table 2-8.  Simple Correlation Coefficients Between
                  Alternative Measures of Long-term Exposure
                  to Ozone
03(1) 03(2)
03(1) 1.00 .783
03(2) 1.00
03(3)
03(4)
03(5)
03(3)
.886
.549
1.00


03(4)
.700
.469
.664
1.00
*

03(5)
.697
.789
.451
.593
1.00
      A  number of  important  assumptions  were  made in  compiling the

multi-year  average  pollution  concentrations   (for  1974-79).    When

calculating the 1974-79 annual summary statistics, different years were

given  the  same  weight  even  though there  were  often  more  readings

available for a particular monitor in one year than another.  Also, the

differences  in the  number  of  observations  across  all  the  monitors

within 20 miles were not weighted in compiling  summary statistics for

any  one  year.   Differences  across  years in  the quality of monitoring

itself have  not  been  accounted  for here.   These  problems  combine to

-------
                                 2-42






make  it  difficult to  derive any  single  number that  characterizes an




individual's six-year exposure to air pollution.  Of course, any single




measure of  long-term exposure is  bound to be  imperfect.   However, we




believe the approach taken here—which uses as many as six years of air




pollution data—is a step in the right direction, one which we have not




seen  elsewhere.    Because  the  chronic effects  of  air pollution on




individuals' health are an area  of  concern,  further efforts to measure




long-term  exposure to  pollution  are  important.   As  the   quality of




monitoring data continues to improve, the payoffs from such a task will




be large.








      2.3.^  Description of the Pollution Variables




      A large  number of different  air  pollution measures  or variables




were  used  in  this report so  it  may  be  helpful  to describe  in  some




detail  the  system  we  used  to  name  these   variables.    The  basic




descriptors for the pollution variables are the prefixes:




                         03  —  Ozone



                         SP  —  Total suspended particulates




                         S2  —  Sulfur dioxide




                         SU  —  Sulfates




                         N2  —  Nitrogen dioxide




                         CO  —  Carbon monoxide




These descriptors  are  always the first two  characters  in  the names of




the pollution variables.  Thus,  any variable beginning with SP is  some




measure  of total  suspended  particulates.    Similarly,  the  prefix N2

-------
                                 2-43


signifies some  measure  of nitrogen  dioxide.   To  indicate  whether the

appropriate  pollution  measure  comes  from  each  individual's  first

nearest, second  nearest,  ..., tenth  nearest  monitor for each  type of

pollutant,  the  third and fourth characters  in  the variable name are

indexed from  01 to 10.   The fifth  character in the variable  name is

either R, L,  M, or  A,   Respectively, this signifies  data  specific to

the   two-week   reference   period;   the   two-week  period  immediately

preceeding the reference period;  the average of these adjacent two-week

periods  (weighted  by the  number of  observations  available  from each);

or the entire  year.   Combined, the  sixth and seventh characters range

from  01  to 06.    These  six  suffixes  describe the  particular  summary

measures for  each  pollutant, at each  monitor, and  over  each possible

summary period, as follows:

               01  Mean of daily maximum for summary
                   period

               02  Number of days for which at least one
                   daily observation was available

               03  Highest hourly or daily reading

               04  Mean hourly or daily concentration

               05  Number of hours  or days in a summary
                   period in which reading exceeded some
                   specified value

               06  Number of hourly or daily
                   observations available during the
                   summary period

where hours/days refers to the maximum frequency of data availability.

      Therefore,  for  example,  0301R01  signifies  the  average  daily

one-hour maximum of  ozone  during the two-week reference  period at the

-------
monitor  nearest   each   individual's   home:   N207L04  is  r.ean  hourly




concentration of nitrogen dioxide during the two-week period preceeding




the two-week  reference  period  at  the  seventh  nearest monitor  to the




individual's  home;   and   SPOUA05  is   the   highest  total  suspended




particulate reading recorded at the fourth nearest monitor.




      Measures  where  the  third  character  is  "N"   are  those measures




obtained from  the monitor nearest  the individual for  which data were




available during the summary period.   In such cases, one of  3, L, M, or




A is the fourth character, with the same meaning as  above, and 01,  ...,




06 again index  the type  of summary measure, but are now located in the




fifth  and  sixth  character  positions of  the  variable  names.    For



example.   03NH01  is   the  average   daily  one-hour   maximum  ozone




concentration for  the individual's two-week  reference period from the




nearest monitor  for which  such data exist,  while 03NL04 is  the average




hourly  concentration  of ozone  for the  two-week  period preceeding the




individual's  two-week reference  period from  the nearest  monitor for




which such data are available.




      Not listed  above are the variables summarized from the NADB time




series  data  described in  the  preceeding  section.   We identify these




variables  in  the  following way.   Pollutants are  again indexed by 03,




SP, S2,  CO,  or N2  (sulfate data  were not available for 19711.1979 from




the NADB system tape  we  used).  These identifiers occupy the first two




character  positions  of  the  variable  names.   "AN"  in the third and




fourth character positions means that  the variable is  a six-year annual




summary  derived using the methodology  described  in  Chapter 2.  "GEAV"

-------
                                 2-45






and  "ARAV"  in  the   fifth   through  eighth  character  positions  mean



geometric mean and arithmetic mean,  respectively.   "MAX2" in character




positions  five  through  eight indicates the  second  highest reading for




the year,  averaged in the manner  described above.   "GENU," "ARNU," and




"M2NU" represent  the  number of annual  summary data  points  over which




the geometric  mean,  arithmetic  mean, and  second maximum  averages were




calculated.




      For  example,   03ANARAV   is   the   six-year   average  of  annual




arithmetic  average hourly   ozone  concentrations;   if  03ANARNU=5,  the




average was  calculated using data  from  five  years.   Variables prefixed




by 03, SP, S2,  N2, or CO, with  the third and fourth characters indexed



over  74,  ...,   79  are  the   measures of  data   completeness.    In such




measures,  the fifth through  eighth character positions  in the name are




"NUGM,"  "NUAM,"  or   "NUM2," signifying,  respectively  the   number  of




monitor values  satisfying EPA and  our  data quality criteria  (described




above) for geometric means,   arithmetic means, and second annual maxima.




Thus,  for  example, SP77NUGMr3  means that there  were  data  from three




monitors   satisfying  completeness  and  proximity  criteria  for  the




geometric mean  of  suspended  particulates  in  1977.   Note that geometric




averages  were  used  exclusively  for  suspended  particulates,  while




arithmetic averages  were used  only for  ozone, sulfur  dioxide, carbon




monoxide, and nitrogen dioxide.




      We  recognize  the  complexity  of  our   system   for  naming  air




pollution  variables.   It is necessitated by our desire  to  experiment




with  a  number  of  different   measures  of  exposure   to  the  various

-------
pollutants.  If  it  is  any consolation to  the  discouraged reader, most




of  the  analysis  below  uses only  one  or  two  measures  of  each  air




pollutant.   In  the  analysis  of  acute  morbidity  in  Chapter  4,  we




typically  used  the  nearest  monitor  for  which  "complete"  data were




available.   Thus,  the  variables  almost  always look like  03NR01, ...,




CONR01.    In  the  analysis  of  chronic  morbidity, we  used air pollution




data averaged over all  of 1979,  or multi-year averaged data.  Thus, the




pollution measures look like 03NA01, ..., CONA01 (for the 1979 data) or




03ANARAV,  ..., COANARAV.   In actuality,  it  is  not  overly difficult to




identify the air pollution variables being used.




      2.3.5  Matching Individuals to Monitors



      The algorithm used to match air pollution data to the individuals




in  the  Health Interview  Survey  and  in  our final  data set  used  the




census tract in  which individuals reside  as  the focal point.  It takes




a two-step  approach  to the  matching  process.   In the  first step, the




geographic  coordinates   of  the  centroids  of  each   census  tract  are




calculated  using the  concept of  oriented  polygons.    This procedure




allows  for  the   fact   that   miles  per  degree  longitude   depends  on




latitude.   The  data  used in this  step are from the  Census Bureau's




Urban Atlas  Files.   Importantly,  they are  only available  for  census




tracts within  Standard Metropolitan  Statistical  Areas  (SMSAs).   This




meant that the matching procedure could only  be  used for residents of




SMSAs.    The  implications  of  this  restriction  for  possible  sample




selection bias are discussed in a later section of this chapter.

-------
                                 2-47






      Once   the   geographical  coordinates   of  the   centroids   were




determined, the second  step  was  a relatively straightforward  one.   It




involved  "sweeping  out"  from  the  centroid  of  a  census  tract  and




identifying  the  nearest  ten  pollution  monitors   for  each of  ozone,




suspended  particulates, sulfur  dioxide, nitrogen dioxide, sulfate ion,




carbon monoxide,  total oxides of  nitrogen,  and lead.   This step made




use  of  EPA  data  on  the  geographic  coordinates  of  all  air pollution




monitors, contained in their Directory  of Air Quality Monitoring Sites.




Using  this  sweeping  algorithm  and  the  unique  identifiers  of  each




monitoring  site  described above,  the  distances  in  miles  from  census




tract centroid to  each monitor were  established  for  each qualifying




individual  in  the  dataset.    Having   thus  matched  individuals  to




monitors,  it  was  then - straightforward  to assign  them  the  appropriate




air pollution data from the SAROAD system.




      Throughout  the  rest  of this  report  we  refer to  the  monitors




"nearest  individual's  homes."    We  do  so  only  for  convenience  of




exposition,  however;  this discussion of  matching  should  make  it clear




that  all  individuals  are assigned  data  from  monitors  nearest  (then




second nearest, and so on)  the  centroid of  the  census tract  in which




they lived at the time of the 1979 HIS.









2.U  Meteorological Data




The  meteorological  data  used in  the following  chapters derived from




data collected and published by the National  Climatic Center, National

-------
                                 2-48






Oceanic and Atmospheric Administration (NOAA), United States Department




of Commerce.  All data used were specific to 1979.




      Two NOAA data bases were  used  to  generate  data for our analysis:




monthly and  annual  values  from the Local  Climatological  Data—Annual




Summaries for 1979;  and daily values from the WBAN S"™*ary o£ Day, Deck




3^5, for  1979.   From the former, monthly  and annual values summarized




from data  recorded  at  the  weather  stations  that constitute  the NOAA




monitoring  network  were  available  for  average  temperature,  relative




humidity for the hour closest to noon, average windspeed, precipitation




in  inches,  snowfall,  heating  degree  days,  and  cooling  degree days.




From the latter  data,  the  daily values  of  interest for purpose of this




study   were  maximum   and   minimum   24-hour   temperature,   24-hour




precipitation, 24-hour snowfall, 24-hour peak instantaneous wind speed,




and  presence  or  absence  during the  24-hours peak  instantaneous wind




speed,   and  presence  or  absence  during the  24-hours  period  of  the




following:    smoke/haze,  rain,  or  snow;  and  percent  of  possible




sunshine.   All  told,  this  gave  sixteen  measures   of  climatological




influences at the weather station level, although the overlaps (e.g.,




monthly versus daily precipitation readings) are obvious.




      Due to the nature of the questions regarding health status in the




Health Interview Survey, data recorded in monthly and daily frequencies




were   inappropriate   to    characterize   individuals'   climatological




exposure.  As explained above in the detailed description of the Health




Interview  Survey,  individual  responses  were specific  to  two-week,




six-month,  or annual  recall periods.   Annual   data  are  suitable  to

-------
                                 2-49






characterize climatological exposure when one is exploring the possible




determinants  of  chronic  morbidity.    However,  when  examining  acute




morbidity during a particular two-week period (as collected in the HIS)




neither monthly nor daily data are ideally suited to the task.  Rather,



data measured at daily and monthly frequencies must be transformed into




measures  representative  of  exposure  to  heat,  precipitation,  etc.,




during  the  two-week recall  period  specific to  an  individual's Health




Interview  Survey date.   The  description of  the methodology  used to




perform this task follows.




      For  the  monthly  data,  a  simple weighted-averaging  process  was




used.  The weights were  the number of days in the individual's two-week



recall  period  falling  in  the  month or months  spanned by the two-week




recall  period  divided   by  fourteen.   For example,  if  the  variable of




interest is  X  and the  recall  period  was  March  31  - April 13, then the




average, Y,  was calculated as  Y = ^1/1^)*xMar  +  (13/110*XA  ].   For



recall  periods  falling  entirely within one calendar month,  the monthly




value  of   the  variable  was  used  as  the  two-week  proxy.    Variables




calculated  with  this  averaging method  were  suffixed by  "RF",  e.g.,




PRECIPRF represents  the averaged  monthly precipitation.   Annual data



were suffixed by "AN".



      Data available on  a  daily basis were aggregated to form two-week




measures.    Four   prefixes   were   used  to  describe  the   statistics



calculated from the daily climatological  data:   "NU" signifies number




of  days over  the two  week period  for  which  the  relevant  data were




available;   "MX"  signifies  the  maximum  recorded  value  during  the

-------
                                 2-50






two-week  period  for the  variable  in  question;   "MN"  signifies  the




minimum recorded value;  and "AV" signifies the mean calculated over all




available observations.   An example should serve to clarify things.  If




the  daily   values  for   maximum   temperature  recorded   during  the




individual's         two-week         recall         period         were




(75,80,85,n/a,n/a,85,80,75,n/a,n/a,80,80,80,n/a),  then  the  following




variables would assume the values listed:




                             NUMAXTMP	9




                             MXMAXTMP	85




                             MNMAXTMP	75.




                             AVMAX7MP	SO




Obviously,  not all  of   the  four  statistics   described  above will  be




applicable in all cases.




      The  variables   derived  from  the   daily  meteorological  data  and




their descriptions are as follows:




      Daily Maximum Temperature




         MXMAXTMP, AVMAXTMP,  NUMAXTMP




      Daily Minimum Temperature




         MNMINTMP, AVMINTMP,  NUMINTMP




      Daily Peak Gust




         MXPKGUST, AVPKGUST,  NUPKGUST




      Daily Presence/Absence of_ Rain (1,0)




         AVRAINYN, NURAINYN




      Daily Presence/Absence of_ Snow (1,0)




         AVSNOWYN, NUSNOWYN

-------
                                 2-51






      Daily Presence/Absence of_ Smoke or Haze (1.0)




         AVSMOKHZ, NUSMOKHZ




      Daily Precipitation _in Inches




         MXPRECIP, AVPRECIP, NUPRECIP




      Daily Snowfall i£ Inches




         MXSNOWFL, AVSNOWFL, NUSNOWFL




      D_ai_l_y Percent of_ Possible Sunshine




      MXPCSUNP, AVPCSUNP, NUPCSUNP




      The  process  for  matching  weather stations  to  individuals  was




based on the Standard Metropolitan Statistical Area (SMSA) in which the




individual  resides.   For each SMSA  in the data set,  one and only one




weather station was  assigned.  In most cases, only one weather station




was located in or near the SMSA  and in these cases,  the data recorded




at that station were assigned to  all  residents  of  the SMSA.   In a few




cases, more than  one station  was available  (e.g., for New York City,




there were data from Central Park, LaGuardia Airport, and JFK Airport),




in which  case  individuals were assigned data from  the nearest weather




station.









2.5  Other Data




      In   addition  to   the   data   described   above   pertaining  to




individuals' health  and  socioeconomic status  and their exposure to air




pollution and  meteorological  factors,  there are several other types of




data  that  deserve  mention here.    Certain  of  these  data  describe




potential  determinants  of  health that have been  included  in previous

-------
                                 2-52






epidemiological  investigations,  although  not  to our  knowledge  in a




study of  this  size  and scope.  Other  data measure variables that have




not before been used to analyze acute and chronic morbidity,








      2.5.1  Pollen




      In order to isolate the  health effects  associated with ozone  and




other  pollutants,  it  is   necessary  to   control for  as  many  other




influential  factors   as   possible.     The   exclusion  of  important




explanatory variables  leads to biased estimates  of  the  parameters on




the  included  variables, unless  all  (vector)  elements  of  the  set of




excluded  variables  are  orthogonal  to all  (vector) elements of the  set




of included variables.   In any event,  the constant  term,  if included,




will be biased.




      The Health  Interview  Survey  and the  air  pollution  data from  the




Environmental  Protection  Agency's  SARDAD  system,   along  with  other




relevant data, provide information about many of  the factors that mi£,ht



affect  acute  and chronic  morbidity.   However,  none of  these sources




provides  data  on  one  potentially  important  determinant  of  health




status—airborne  concentration of  pollens  in the geographic areas of




interest.     Since   observed  ill   health—particularly  respiratory




discomfort—may be  due not to  air pollutants,  but  rather  to airborne




allergens,  it  is  clear   that  controlling  for  such   influences  is




important if reliable parameter estimates are desired.




      Unfortunately,  pollen  data   are  neither  widely  available   nor




highly  reliable.   To  our knowledge, there  exist no  federal government

-------
                                 2-53






sources.  After investigating a  number  of  possible sources, we settled




on  data amassed  by  the Pollen  and Mold Committee  of  the  American




Academy of Allergy  (AAA).   In its  annual report, AAA publishes data on




a broad array of pollens monitored at various sites nationwide.




      There are several  problems with these data.   First, not all AAA




sampling stations collect data on  all kinds  of pollens.  Some stations




collect only  ragweed  data,  while others monitor  up  to forty different




pollen  types.     Second,   different  stations  employ  different—and




noncomparable—types  of  collection devices.   The  most common of these




are the Durham Sampler (DS) and the Intermittent Rotorod (IR), although




others  exist.   Some  of  the methods use planar  measures per unit time




(e.g.,  particles/cm /2U  hours), while  others use  volumetric measures




(e.g.,  particles/m-).    There  are  no  generally acceptable  methods  of




conversion  between   these   two  measurement   types,   not  even  at  a




"rule-of-thumb"  level.    This  limits  the number of  sites  for  which




comparable data  are available.   Third,  not  all  stations  collect data




every  day of  each  month,   i.e.,   figures  are  presented  monthly  for




aggregate particles collected and total collection days.  The best that




can  be done  here—absent  other  information—is to  assume  that  the




sampling is random,  that the average daily  counts for the sample days




in fact represent the overall averages.




      However, we decided the AAA  data  could be useful  in our efforts.




Twenty-five stations in the metropolitan areas surveyed  by NCHS collect




data using the DS method.  This guaranteed a large number of individual




observations while  simultaneously  allowing the data  to be derived from

-------
                                 2-5U






a single sampling method, thus ensuring comparability.  We used DS data



in our statistical work.




      Although we  would  prefer  to  have  data on  all pollens  in each




SMSA, several medical experts suggested that a ragweed-only index would




be an acceptable measure  of  pollen-related  maladies,  indeed by far the




best single measure. ^




      In  the  analysis   below,  we   test   the  seriousness  the  data




limitations  we   face   by  estimating  models  with   a   subsample  of




individuals  for  whom  there  exist  pollen  data  and  comparing  the




estimated coefficients with those from regressions where no pollen data



are included.








      2.5.2  Indoor Air Pollution




      There   is   a  growing  awareness  among   environmental  health



                                                       1 ?
scientists of the risks  posed by indoor air pollution. J Concentration




gradients  exist  for  a  variety  of  air  pollutants  between  indoor




(residences,  for  example)  and   ambient  air.    Carbon  monoxide  and



suspended particulates are likely  to range  higher in homes occupied by




cigarette smokers; radon gas, which is trapped indoors, is likely to be




high in certain  sections  of  the  country;  nitrogen dioxide is likely to




be higher in homes that use gas stoves for cooking; while ozone,- sulfur




dioxide, and sulfate particles are likely to be higher in ambient air.




      The  magnitude  of  the gradient  reflects  the  strength of  the




pollutant  sources   and  the  ventilatory   capacity  of   the  indoor




environment;   increased  ventilation  acts  to  reduce  the  gradient.   In

-------
                                 2-55






buildings with  potent  sources—whether inefficient  space  heaters, gas




ovens,  or  smokers—and  low  rates  of  ventilation  (tightly  sealed




buildings),  peak indoor  concentrations  can  exceed  those experienced




outdoors, and even exceed the NAAQSs.  Thus, one would like information




about  such  indoor exposures  in order  to isolate health  effects  that




might  arise  from exposures to  outdoor concentrations.   Unfortunately,




no information  about  cooking fuels is  elicited  in  the  HIS, nor is any




other  information that  could  provide  some guide to indoor exposures to




air pollution, apart from the information in the smoking supplement.




       Although no individual data are  available, it is possible to draw




some  rough inferences  about  likely indoor exposures to carbon monoxide




and  nitrogen  dioxide  from  natural  gas  for  the  individuals  in our




sample.   Using the  Provisional  Estimates _of_ Social^  Economic, and




Housing  Characteristics  of the  1980  Census  of  Population and Housing,




data  were collected on  the  number  of  occupied housing units in each of




the  thirty-eight SMSAs  in  the United  States   with  greater  than one




million  persons  that  use natural gas  (either piped  or  bottled) as the




primary  cooking fuel.   We  converted   these to  percentages of occupied




housing  units  and,  in  some  of  the   regressions  below,  assigned all




households in a  particular SMSA  the average for  that SMSA.  This ranged




from  a low of  U percent  in the Seattle-Everett SMSA to  a high of 88




percent in the New York-New Jersey SMSA.  The figure can be interpreted




as  the probability  that  the  household  cooked  with natural  gas, and




hence  the  probability that its  members were exposed to possibly high




levels  of  carbon  monoxide  and  nitrogen  dioxide indoors.   Individual

-------
                                 2-56






data comparable  to  those on health  and socioeconomic status  would be




preferred.   However,  the aggregate  data from  the census  permitted a




crude control for indoor exposures.








      2.5.3.  Paid Sick Leave



      One of  the major  interests  in this  report is  the  relationship




between  air pollution  and  time  lost  from work,  other factors  held




constant.    One  of  these  other  factors  which  may  influence  work




loss—and hence  which  should  be  controlled—is  the  availability and




amount of paid sick leave that a worker- enjoys.   Two individuals may be




identical in every respect studied here but for the amount of paid sick




leave  they  are  granted each  year;   if  so,  more  work loss  might be




expected  from  the individual with more  paid  sick leave  because a day




lost from work  will not cost him  (or her)  as much  as  it will cost the




other worker  (assuming  that  the limits are binding).   The presence of




paid  sick leave may  therefore lower the  "threshold"  of  illness  that




must be reached before a worker stays home from the job.




      An  alternative  hypothesis might be  advanced,  however.  Paid sick




leave may enable  workers  to  stay  home and rest  during the early stages




of an illness, thus reducing their chances of more serious illness that




could lead  to  prolonged work  loss.   If  so, workers with more paid sick




leave might  be  expected to miss less work  each year,  ceteris par i bus,




than workers with less.   At  any rate, paid sick leave should play some




role in explaining the amount of work loss.

-------
                                 2-57






      No questions  were  asked on  the  1979 HIS about  paid sick leave.




However, in  1974  a  supplement on conditions of current  employment was




included in the HIS in which each working respondent was asked how many




days  of "wage loss  coverage  (sick  leave,  etc.)..."  they  received




annually.     We  averaged the  reported  paid  sick  leave  for  1974  by




occupation and  industrial category.   Then,  assuming that  the combined




industry-occupational averages  have  remained  the  same  since  1974,  we




assigned the same  number of  paid sick  leave  days to  individuals  in




various industries  and occupations for  1979.  While it would have been




preferable if  the  1979  HIS  had  contained a question  about  paid sick




leave,  using  the  1974   data  from  the  same  survey  seemed  to  be  a




reasonable compromise.   It is unlikely that the number of days of paid




sick  leave will have changed  much since  1974, and  even more unlikely




that  the relative rankings between  industries  and/or  occupations will




have changed.









      2.5.4  The Pollutant Standard Index




      Individuals must  venture outside  during  polluted  periods if one




is to observe  a relationship  between either acute  or chronic morbidity




and  air pollution.    Yet it  is  possible that  if  individuals  are made




aware  of periods  of poor  air  quality,  they  would forego  or reduce




outdoor  activities  during  these periods.    If   such  behavior  were




widespread,  the observed relationship between  acute morbidity and air




pollution might be negative, rather than as  expected.

-------
                                 2-58


      Conversely,  information  or perceptions  about current  levels of

air quality might  have a  different  effect.   Individuals aware that the

air quality during a  particular  period  was  very poor might become  even

more aware  than usual of respiratory  and  other ailments  and  show an

expectational   effect.      Since  acute  and   chronic   morbidity  is

self-diagnosed  in the  Health Interview Survey,  this  possibility is

pertinent to the present  study.

      To  explore  the  possible  role  played by information  about air

quality  conditions,   we   obtained  from  the  Environmental  Protection

Agency  a summary  of  the  metropolitan  areas  in the United • States in

which  air  quality conditions were regularly  reported  in  1979.   In

particular, those areas reporting the Pollutant Standard  Index  (PSI)—a

convenient summary measure of air quality—were identified. 5  Adoption

of  the PSI  was used  to  proxy  the availability of  ambient  pollution

information.  A  dummy variable was  created  for each area taking on the

value  1 if the PSI was reported  in  1979, and the value 0  if not.

      Quoting from an EPA report:


          The PSI is presently being reported in 94 urbanized
          areas, accounting  for  89  million  people.   Of  these
          94  urban areas, 39 areas  have populations greater
          than  or  equalling  500,000;   39  have  populations
          ranging  from   200,000  to  500,000;   and  16   have
          populations less than  200,000.

The PSI  adoption  dummy, PSIY1NO, was  matched to the adults data set by

SMSA identifier.   In  the  sample  of  1U,UU1 adults,  12,615  (87.455)  lived

in areas where the PSI was reported while 1,826  (12.6/f)  lived in areas

-------
                                 2-59






where  it  was  not.   The  EPA report  used  by  RFF  to  construct  this




variable is available upon request.









      2.5.5  Availability of Medical Care




      In  addition  to  the  determinants   discussed   above,  acute  and




chronic morbidity may depend  in  part  on access  to medical care.  Those




who  can receive  medical  attention  quickly  and  inexpensively  can be




expected to enjoy  better  health  than  those for  whom medical care is an




expensive and time-consuming activity.  Unfortunately, the HIS contains




no  information  about individual  access to  medical  care.   Rather, it




concerns itself with acute and chronic illness,  physician visits and so




forth, without indicating  the lengths  to  which  an individual had to go




to obtain that care.




      In an effort to get some idea of its possible effect on acute and




chronic morbidity, we included in  our  data base two measures of access




to medical  care:   the  per  capita  numb.r of  physicians  in  the  SMSA in




which an individual  lives;  and the per  capita  number of hospital beds




in  that SMSA.   These  data  were derived  from  the U.S.  Department of




Commerce,  National  Technical Information  Service  (NTIS),  Bureau of




Health Manpower Area Resource File  (ARF).   The  physicians and hospital




data  were  derived, respectively,  from the  health manpower  and health




facilities sections of  this  file,  while SMSA population was drawn from




the population section of the same file.

-------
                                 2-60


2.6  Problems of_ Sample Selection

      When  using  data  derived  from  other  than  purely  randomised

sampling  procedures,  one must  understand the  implications of  and be

prepared to correct for sample  selection  bias.   This is most important

when one  goal of  research  is  to  extrapolate results  from estimation

based on  samples  to a  group  not represented in  the observations over

which the  parameters were  estimated.   The problem  has  been summarized

as follows:

          Sample selection bias may arise in practice for two
          reasons.  First, there may be self selection by the
          individuals  or  data  units  being  investigated.
          Second,  sample  selection decisions by  analysts or
          data processors operate in much the same fashion as
          self selection.

          There  are many  examples of self  selection  bias.
          One observes  market wages for  working  women whose
          market wage  exceeds their home wage  at zero hours
          of  work.   Similarly,  one observes  wages  for union
          members  who  found  their nonunion  alternative less
          desirable.    The  wages  of  migrants   do  not,  in
          general,  afford   a  reliable  estimate   of   what
          nonmigrants  would  have  earned  had they  migrated.
          The earnings  of  manpower trainees  do  not estimate
          the earnings that nontrainees would have earned had
          they opted  to become  trainees.   In  each  of  these
          examples, wage  or  earnings  functions  estimated on
          selected  samples  do  not,  in  general,  estimate
          population  (i.e.,  random  sample)  wage functions.
          Comparisons of the wages of migrants with  the wages
          of  nonmigrants (or trainee earnings with nontrainee
          earnings, etc.) result in  a  biased estimate of the
          effect   of   a  random  "treatment"  of  migration,
          manpower training,  or unionism.

          Data may also  be  nonrandomly selected because of
          decisions  taken  by data  analysts.   In studies of
          panel   data,  it   is   common   to   use   "intact"
          observations.  For example, stability of the family
          unit is  often imposed as  a requirement  for entry
          into a  sample  for  analysis.   In  studies  of life
          cycle fertility and manpower training  experiments,

-------
                                 2-61
          it  is  common  practice  to  analyze  observations
          followed for  the  full length of  the  sample,  i.e.,
          to  drop  attriters   from  the   analysis.     Such
          procedures  have the  same  effect  on  the  structural
          estimates  as  self  selection:   fitted  regression
          functions  confound  the  behavioral  parameters  of
          interest   with    parameters    of   the    function
          determining  the  probability of  entrance  into  the
          sample.1'

      This  researcher  then  summarized  the  problem:   "The  critical

question is 'why are  the data missing?'"

      In our case, the  discussion above has demonstrated  "why the data

are missing."   The obvious  follow-up  is  how to deal with the missing

data problem.

      In our study if  an adult was given  a supplemental  questionnaire

about smoking, he was  eligible  for  inclusion in our final sample if we

could match him to at least one air pollution monitor of any kind.  But

this  required   that  the  individual  live   in  a Standard  Metropolitan

Statistical Area  (SMSA).   Not  all residents  living within  SMSAs  are

included  in our  final  sample, but  no  one living  outside  SMSAs  is

included.   The  following  describes   our  working  hypotheses for  the

sampling aspects of this project.

      Let

                S = final sample

                M = set of individuals receiving the smoking survey

                C = the set of children (aged 16 and under)

                P = the set to whom monitors could be matched

                V = the set of individuals surveyed living in SMSAs

                R = the set of individuals surveyed not  in SMSAs

-------
                                 2-62
                T = the overall sample ( #T = 110,530)




                CQ(A) = the complement of Q in the universe A
We then have
                S = {  i i  IE (C a P) U (CC(T) .1 P -1 M)}      (1)




                Pr( i  eS i  i £R ) = 0                        (2)




                Pr( ieV I  ieC U (CC(T)  a M )) = f(Xu;B1)  (3)




                Pr( ieS 1  i£V ) = Pr( i£S i  i^T ) -




                   Pr( i£S I  isR ) = g(X2i;S?)
Identity  M)  is merely  definitional;  (2)  states  that  S  contains no




individual living outside an SMSA; (3) states that the probability  that




an individual  resides  in an SMSA given that the  individual  is in the




entire sample  (and given  that  a  smoking  survey was  administered, if




adult) depends  on  a set of  characteristics of  the  individual, X    a




parameter vector,  B. ,  and  a  functional relationship  f(  );  (4) states




that  the  probability  than an individual  is in  the  final  sample given




that  the  individual  resides  in  the  SMSA  depends  on  a   (perhaps




different) set  of  individual  characteristics,  X-., a parameter vector,




£?, and a functional relationship g(  ).








      The  first question  to  explore,  then,  relates  to  (3).   What




factors determine whether or not  an  individual  resides in an SMSA?  We




want to consider

-------
                                 2-63


           f: Xu __> [ 0, 1 ] , such that

           Pr( ieV  I ±cC  U (Cc(T)n  M)) = f(Xu; 6^         (5)

We will now explore a somewhat restricted form of this hypothesis:

             Pr( lev i i£ CC(T) n  M) = f (Xu; «.,)              (6)

Our  approach  is as  follows.   We  assign to  each adult  receiving the

smoking  survey  (i.e. all  i £ M) a  binary  (0,1) variable  called SMSA

defined as

             SMSA = 1  iff ie V                                (7)

                  = 0  iff i £ R

We then consider a model of the form

             Pr(SMSA1 = 1) = ?(XU 2,)                         (8)

where F(   )  is a distribution  function.   Our task  is  to estimate the

parameter  vector  £    given X.   We  will assume  that  ?(   )  follows a

logistic distribution, so that

             ?r(SMSA._ = 1) = (1  + exp(-X1:L E^)"1              (9)

The X, vector has the following elements
             NE	Northeast region dummy
             NC	Northcentral region dummy
             SO	South region dummy
             SEX	Sex, 1 if male, 0 if female
             AGE	Age in years
             RACE	Race dummy, 1 if white, 0 else
             MAR	Marital status , 1 if married, 0 else
             EDDC	Years of schooling
             INC	Pseudo-continuous measure of income, in $
             CONS	Vector of unity elements

      The 6 vector estimated by maximum likelihood estimation of (9) is

as follows, with asymptotic t-scores in parentheses.

-------
                                 2-64
               NE       .115 (2.25)
               NC      -.607 (13.20)
               SO      -.977 (22.20)
               SEX     -.057 (1.90)
               AGE     .0027 (3.14)
               RACE   -1.033 (18.78)
               MAR     -.453 (13.73)
               EDUC     .043 (8.60)
               INC  .0000385 (22.65)
               CONS    1.257 (13.23)

      All estimated parameters  are  significant  at above 95* except for

the parameter  on SEX, with most  parameters significant  at well above

99?.   For example, these results  indicate that people  living in the

northeast part  of  the country  are  more likely than  those  in the west

(the excluded  dummy)  to  live  in SMSAs.   Similarly,  SMSA residents are

more likely to be elderly, nonwhite, unmarried, more educated, and have

higher incomes than those outside SMSAs.

      A second question pertains  to (4)  above.   That is: given  that an

individual resides in an SMSA, what are the factors influencing  whether

or  not  such  an  individual  will be included  in our final  sample?   We

want to  consider,  given  an attribute  vector X,. and  parameter vector

j--1
~2'
                   g:  X2i —> [ o,1 ], such that

                   ?r( ie S I  ie V  ) = g( X2i?  32)                (10)

Again, we look at a restricted version of this,

         Pr(  leS I  i£V n CC(T) n M ) = g( X2i;  &2)              (11)

      Now  we  consider  the  possibility  that   —  in  addition  to  the

elements of  X  — the residence  of an individual within or outside a

central  city  of the  SMSA will  influence  the  probability  of being

-------
                                 2-65


included in  the  final  sample.  Thus, X   equals  X.  plus the additional

element  of  INCITY,  where  INCITY  =  1   if  the  individual  resides  in

central city areas  of  SMSAs and =  0  if  the individual resides outside

central  city areas  of  SMSAs  but  still  within  SMSA boundaries.   The

binary dependent variable is now called INSAMP, defined as

                   INSAMP = 1 iff i £ S

                          = 0 Iff ie Cg(v)                        (12)

Again we posit a logit specification, so that

                   ?r(INSAMPi = 1) = (1  + ex?(-X2i S2))~1         (13)

With  asymptotic  t-scores   in  parentheses,   the  maximum  likelihood

estimator of  £  yields,


                   NE         .790 (13.33)
                   NC         .680-(11.72)
                   SO         .050 (0.940)
                   INCITY     .637 (14.29)
                   SEX        .109 (2.67)
                   AGE        .015 (11.19)
                   RACE     -.054 (0.819)
                   MAR      -.270 (5.82)
                   EDUC     -.068 (9.30)
                   INC  -.0000138 (5.76)
                   CONS     1.450 (11.33)

Only  the  parameters  estimated  on DUMSO  and  RACE  fail  to  attain

significance  at  the  99?  level.    Clearly,   the   INCITY  effects  are

overwhelmingly significant, indicating that individuals residing in the

central  city areas   of  SMSAs  are more likely  to be selected  into  the

final  sample—other things   equal—than  those  residing  outside  such

areas.  This follows from the tendency for pollution monitors to be

-------
                                 2-66






more  densely bunched  in -central  city  areas  than  in  outlying  areas




within SMSAs.




      (Because  the question  of  the  determinants  of  Pr(INCITYrl)  is




itself relevant, we estimated a  third logit  model which indicated that




RACE  and  INC, both negatively,  were  the most  significant explanatory




variables as  judged by asymptotic t-scores.   The  full results of this




estimation are available on request.)




      We  have in  effect  analyzed two  conditional  probability models,




first, the probability of living in an SMSA given that a smoking survey




was  administered,  and second,  the  probability  of  being  in  the  final




sample  given SMSA  residence.   Both of  these conditional  probabilities




were  seen  to  be   systematically  related   to   vectors  of  attributes




characterizing the individuals/elements in the relevant universes, and,




therefore, were seen to be representative of nonrandom events.




      The  reason  that  conditional  probabilities  were  analyzed  above




owes to t.ie major motivating factor for this study:  drawing inferences




about larger  groups from  parameters estimated  on subsaiaples.   One such




group,  of course,  is  the  total population,  while  another  group  of




interest is  the total SMSA population.  Results not presented here, but




available  on request,  demonstrate  that  analyzing the  probability  of



being included  in  the  final  sample,  given  that a  smoking  survey was




administered, is  significantly influenced by  individual  socioeconomic




attributes,  as expected.   Thus,  it  seems that in extrapolating results




either  to  the  total  population or  to  an  SMSA  population  one  must




proceed with  caution.

-------
                                 2-67






      To illustrate the  differences  between individuals in "he overall




HIS, those receiving the smoking supplement, and those in the final RFF




sample, we have  calculated  population  summary statistics for the three




groups.   Table  2-9 presents  the  comparisons  for adults,  while Table




2-10 refers  to children.   One  important point is worth noting.  Even




though our sample  selection procedure  is biased towards SMSA residents




(no one outside an SMSA is included in our sample), and  even though our




procedure  has a  (less  serious)  bias  toward  central  city residents




(because of  the  availability of pollution  monitoring data there), the




characteristics  of  our final samples  of adults and  children  are very




similar to those of the  overall population.  Only with  respect to SMSA




and  Central-city   residence   are   there  significant   differences  in




population characteristics.   Our sample  does have slightly more adults




and  children from the  northeast region of  the U.S. than  the  HIS and




slightly fewer from the  south,  but the age, income,  education, health,




labor  market,  and  other characteristics are  quite  similar between the




two samples.




      The distinctions between  the two samples  arise again  in Chapter 5




when we  discuss  the  possible  extrapolations of our  results to larger




populations.

-------
                                     2-68
Table 2-9.  Population Summar:^ Statistics:  Adults
All Adults (HIS) Smoking Survey Final RFF Sample
N

Age

Restricted activity days
(two weeks)
Bed disability days
( two weeks )
Work or school loss days
(two weeks )

Region of residence
Northeast
Northcentral
South
West
Location of residence
SMSA-Central City
SMSA-not Central city
Not in SMS A
Location of residence
SMSA
Non SMSA non-farm
Non SMSA farm
Race
White
Black
Other
Sex
Male
Female
Age
15-44
U5-6H
65-
79,743
Mean (Std. Dev.)
42.54
(18.22)
0.8«7
(2.88)
0.286
(1.55)
0.160
(1.09)
Frequencies (?)

22.64
26.54
32.18
18.6"

28.21
40.54
31.26

68.75
28.47
2.79

88.55
9.84
1.60

46.75
53.25

57.34
27.95
14.71
26,271

42.42
(18.19)
0.851
(2.88)
0.292
(1.57)
0.162
(1.10)


22.61
26.37
32.34
18.68

28.30
40.43
31.27

68.73
28.46
2.81

88.41
9.91
1.68

46.71
53.30

57.57
27.95
14.49
14,UU1

42.98
(18.34)
0.882
(2.93)
0.313
(1.630
0.169
(1.12)


28.40
26.52
25.46
19.63

44.37
55.63
0.00

100.00
0.00
0.00

85.88
12.14
1.98

46.53
53.47

55.37
29.80
14.83

-------
                                     2-69
Table 2-9 (continued)
N
                           All Adults  (HIS)   Smoking  Survey
79,7143
26,271
Final RFF Sample
     14,441
Mean (Std. Dev.)
Marital status
Married, spouse present
Widowed
Never married
Divorced
Separated
Married, spouse absent
Education (years)
None
01-08
09-11
12
13-15
16*
Unknown
Veteran status
Nonveteran
Peace time only
WWI
WWII
Korean War
Vietnam War
Post Vietnam
Don't know
Family income
under $3,000
$3,000-^,999
$5,000-6,999
$7,000-9,999
$10,000-1U,999
$15,000-24,999
$25,000+
Don't know
Living arrangement
Living Alone
Living w/non relatives
Living w/spouse
Living w/relatives-other

63.81
7.62
20.33
5.38
2.16
0.69

0.71
14.00
16.56
37.24
15.69
14.02
1 . 77

79.59
2.55
0.55
6.66
3.21
5.17
0.21
2.05

4.26
5.90
6.84
8.80
15.50
24.77
24.47
9.47

11.57
2.90
63.75
21.78

64.13
7.43
20.21
5.36
2.21
0.65

0.71
13.90
16.62
37.32
16.01
13.96
1.49

80.01
2.43
0.52
6.81
3.07
5.18
0.24
1.76

4.29
5.83
6.81
8.80
15.75
24.73
24.39
9.41

11.40
2.91
64.06
21.63

60.39
7.83
22.87
5.77
2.45
0.70

0.75
13.35
17.01
36.69
16.54
14.02
1.64

79.63
2.43
0.5^
7.47
3.17
4.70
0.21
1.85

4.02
5.75
6.61
8.32
14.90
24.40
25.77
10.24

12.31
3.05
60.3^
24.30

-------
                                     2-70
Table 2-9 (continued)
                           All Adults (HIS)    Smoking Survey   Final RFF Sample
                                79,743           26,271              14,441
                               Mean (Std.  Dev.)

Usual activity
  Usually working                57.23             57.63              57.08
  Keeping house (female)         24.12             23.88              23.22
  Retired-Health (45+ yrs)        2.46              2.61               2.41
  Going to school                 7.85              7.60               8.14
  Something else                  3.16              3.18               3.56
  Retired-other (45 yrs)          4.96              4.93               5.44
  Don't know                      0.22              0.18               0.15

Activity limitations
  Cannot perform usual activity   4.94              5.15               5.24
  Can perform usual activity,     9.24              8.99               8.55
    but limited in amount and
    kind
  Can perform usual activity,     u.39              4.18               4.11
    but limited in outside
    activities
  Not limited or not applicable  81.43             81.69              82.10

Health status
  Excellent                      43.99             44.13              43.39
  Good                           39.54             39.47              40.46
  Fair                           11.97             11.98              11.83
  Poor                            3.82              3.77               3.63
  Unknown                         0.68              0.66               0.69

-------
                                 2-7'
Table 2-10.  Population Summary Statistics:  Children
N

Age

Restricted activity days
(two weeks)
Bed disability days
(two weeks)

Region of residence
Northeast
Northcentral
South
West
Location of residence
SMSA-Central City
SMSA-not Central City
non-SMSA
Location of residence
SMSA
Non SMSA non-farm
Non SMSA farm
Race
White
Black
Other
Sex
Male
Female
Age
Under 6
06-16
Family income
under $3,000
$3,000-4,999
*5, 000-6, 999
$7,000-9,999
All Children (HIS)
30,787
Mean (Std. dev.)
8.49
(4.93)
0.417
(1.62)
0.186
(.920)
Frequencies (*)

20.50
27.33
33.88
18.29

26.14
40.55
33.32

66.68
30.59
2.73

83.63
14.43
1.94

50.99
49.01

31.49
68.51

3.26
4.86
5.83
8.09
Final RFF Sample
15,711

8.61
(4.94)
0.416
(1.61)
0.191
(.944)


25.83
28.71
26.81
18.65

43.88
56.62
0.00

100.00
0.00
0.00

79.19
18.59
2.22

51.05
48.95

30.66
69.34

3.32
5.75
6.15
7.89

-------
                                 2-72
Table 2-10 (continued)
                               ill Children (HIS)
Final RFF Samole
N 30,787
Mean (Std.
$10, 000-14, 999
$15,000-24,999
$25,000+
Don't know
Activity limitations
Cannot perform usual activity
Can perform usual activity,
but limited amount and kind
Can perform usual activity, but
limited in outside activities
Not limited or not applicable
Health status
Excellent
Good
Fair
Poor
Unknown

dev. )
16.42
29.01
23.52
9.02

0.19
1.95

1.85

96.02

60.13
34.33
4.24
0.52
0.78
15,711
14.92
28.31
23.82
9.84

0.20
2.20

1.81

95.78

58.79
33.51
4.37
0.54
0.80

-------
                                 2-73





                               Footnotes






      This algorithm was developed by Professor Raymond Palmquist of



the Department of Economics and Business,  North Carolina State



University.



     National Center for Health Statistics [1981], pp. 42-43.



      National Center for Health Statistics [1981]; National Center for



Health Statistics [1975]; Bureau of the Census [1979]; National Center



for Health Statistics [1970].


     u

      See, for example, Lilienfeld and Lilienfeld [1980], chapters 6,
     ^See, for instance, Higgins [19831.



      See Manning, Newhouse, and Ware [1982].



     7See NCHS [1981], pp. 57-58.


     g

      Surgeon General of the United States [1982].



     9Warner [1978], p. 314.



      U.S. Environmental Protection Agency [1980], p. vii.



      See U.S. Environmental Protection Agency [1978], Johnson [1982].



     "Bernard D. Goldstein, M.D., personal communication, 7/6/82.



     -"National Academy of Sciences [1981].


    1 4
      This can be found at Tape location nos. 529-30 in the 1974 HIS.



      For information about the PSI, see Federal Inter-agency Task



Force on Air Quality Indicators [1976].

-------
     "For details about  this  file,  see  U.S.  Department  of  Ccmerce
[19791.



    17
      See Heckman [1979],  pp.  153-15^.

-------
                                 2-75


                        References  to Chapter 2
Bureau of the Census (acting as Collecting Agent  for U.S.  Public Health
     Service), Health Interview Survey Interviewer"s Manual,  U.S.
     Department of Commerce.  Publication no.  HIS-100 (1979),  1979.

Federal Interagency Task Force on Air Quality Indicators,  _A Recommended
     Air Pollution Index (Washington, B.C.:   U.S. Government  Printing
     Office), 1976.

Heckman, James, "Sample Selection Bias as a Specification  Error,"
     Econometrica, vol. U7 (1979), pp. 153-161.

Higgins, Ian, "What Is an Adverse Health Effect?" Journal  of  the Air
     Pollution Control Association, vol. 33 (1983),  pp. 661-663.

Johnson, Ted, "A Jump Test for Validating Hourly Average  Air  Quality
     Data," paper presented at the Annual Meeting of the  Air  Pollution
     Control Association, New Orleans, June 20-25, 1982.

Lilienfeld, A.M. and D.E. Lilienfeld, Foundations of Epidemiology (New
     York:  Oxford University Press), 1980.

Manning, Willard, Joseph Newhouse, and John Ware, "The Status of Health
     in Demand Estimation; or. Beyond Excellent,  Good, Fair,  Poor,"  in
     Victor Fuchs (ed.) Economic Aspects of Health (Chicago:
     University of Chicago Press)~1982.

National Academy of Sciences, Indoor Pollutants  (Washington,  D.C.:
     National Academy Press), 1981.

National Center for Health Statistics, Health Interview Survey
     Procedure, 1957-197U, Department of Health  and  Human  Services
     Publication no. (HRA) 75-1311, 1975.

National Center for Health Statistics, Current Estimates  from the
     National Health Interview Survey;  United States,~979,  Department
     of Health and "Human Services Publication no. (PHS) 81-1564, 1981.

Surgeon General of the United States, The Health Consequences of_
     Smoking, Cancer, Department of Health and Human Services
     Publication no. (PHS) 82-50179, 1982.

-------
                                 2-76
U.S. Department of Commerce (National  Technical  Information Service),
     "Bureau of Health Manpower Area Resource  File User  Documentation,"
     Publication no. HRP-0901718,  July 1979.

U.S. Environmental Protection Agency,  Screening  Procedures  for  Ambient
     Air Quality Data, EPA-U50/2-78-037 (Research Triangle  Park,  N.C.),
     1978.

U.S. Environmental Protection Agency,  Air Quality Data—1979 Annual
     Statistics, USEPA, Office of  Air  Quality  Planning and  Standards,
     no. EPA-U50A-80-01U,  1980.

Warner, Kenneth, "Possible Increases in The Underreporting  of Cigarette
     Consumption," Journal of_ the  American Statistical Association,
     vol. 73 (1978), pp. 314T3T3T

-------
                            Chapter 3








                           METHODOLOGY








     In  the first  part  of  this  chapter, we  elaborate  on  the  brief




description in Chapter  1  of  our approach to health effects estimation.




First we  discuss  measures of  acute  and  chronic  health status used  in




this  study.   It  is the  variations  in  these  measures  which we  later




attempt  to  explain  using  concentrations  of   ozone   and   other air




pollutants,   socioeconornic    measures,    and    additional    possible




determinants.   These   measures of  health  status  are  the   dependent




variables in the series of statistical estimates  presented below.




     Next  we  identify  and discuss  the  explanatory  variables used  to




explain  variations  in  each   of  these   measures  of   health status.




Variables  used  in one  model  might not be used in another;  typically,



however, a  set  of  "core"  variables  turns  up in  each of the  estimated




models  because  they  are  thought  to  be  important  determinants  of   a




number  of  measures of  health  status.   Included  in this discussion  of




independent variables are  some  a_ priori speculations  about their likely




signs in the regressions below.

-------
                               3-2
     We also  discuss  the samples over which  our models are estimated.


We  often  estimate  a  number  of  different  specifications of  the  same


basic  model  on  one  particular  set  of observations.    It  has  been


observed that  this  can  cause problems when only a subset of  the final


results are reported.   We  go to great lengths in this study, however,


to report a  wide  range  of  results,  or to  make them available to those


interested.    Coupled with  the  impossibility  of having  used "fresh"


observations  for  every  alternative  specification, we  consider this  to


be a satisfactory justification for the approach we have employed.


     Finally, we discuss briefly the estimating techniques used in  this


report  to  identify  possible relationships between  the  dependent  and


independent* variables.  The choice of  a particular technique is seen  to
                                            •

depend  in  large  part   on   the  nature  of  the  health  measure  being


examined.   In some cases fairly simple  statistical  techniques such  as


ordinary least squares   (OLS)  are  used.   In  other  cases more complex


approaches have been adopted.





3.1  Measures of_ Health Status;  Choice of_ the Dependent

     Variables


     We have  declined in  this study to measure health status  using the


"excellent,  good,  fair, or  poor"  (SGF?) descriptors.   First, any one


individual might describe health status in one way at one  point in  time


and  another  way  at some  later  point  even though  a  more "objective"


measure  of  health  status  might   indicate  identical   states.     More


importantly,  wherever  possible  in  this  study  we  are  interested  in

-------
                               3-3
establishing quantitative  relationships  between ozone  and measures of




health  status   that  might   lend   themselves  to   valuation.     The




distinctions  possible  using  the  "EGFP"   scale  do  not   easily  lend




themselves to such valuation.




     Well  suited  to this  purpose are other  measures  of  acute health




status reported in  the HIS.   It  will  be recalled that each respondent




was asked  the  number of "restricted activity  days"  (RADs), "work loss




(or school loss) days" (WLDs or SLDs), and "bed disability  days" (BDDs)




due to illness during the two weeks immediately preceding the interview




date.  The number  of each  can vary  from zero  to a maximum of fourteen




during  the  two-week reference  period.    Because  these  measures  are




quantitative and  vary continuously,  and because they  are amenable to




economic valuation,  they  qualify  as  excellent  dependent variables.  We




make much  use of these measures in the analysis below.




     As mentioned earlier, each time an acuta health episode occurs—be




it a RAD,  a  WLD, or  a  BDD—the cause or condition giving rise co it is




ascertained  and coded  according  to  the International Classification of




Diseases.  Thus, during  a particular individual's reference period, he




or she may have one  BDD  due to an acute respiratory condition (asthma,




bronchitis,  etc.),   one  BDD  due to  an  upset stomach  or  some  other




gastrointestinal problem,  and one BDD  because of a  sprained  ankle or




other  accidental  injury.   In other  words, RADs,  WLDs, and  BDDs  are




broken down  by specific types of injuries or diseases.




     Even  though respiratory health effects are those most  likely to be




associated with  air pollution, we  begin our  analyses  below using the




number of  BDDs, WLDs (or SLDs) and RADs during the reference period due

-------
                               3-4
to all causes  (including accidents).   We  do  this  for several reasons.




First, as  a practical  matter,  respiratory disease  typically accounts




for a substantial fraction of the total number of 3DDs, WLDs, SLDs, and




RADs from all causes each year  in  the United  States.  In the 1979 HIS,




for  example,  the percentages are  47, 37,  57,  and  40,  respectively.




Second, some evidence suggests  that ozone  and other air pollutants may




be related to eye irritation, cardiovascular disease, susceptibility to




infectious disease, and other ailments as well as respiratory disease.




For  this  reason, it is important to  look beyond  respiratory disease




alone.




     Following the  analysis of  RADs,  WLDs,  SLDs  and HDDs due  to all



causes,  we  turn  our   attention  specifically  to  respiratory-related




impairment.  The overall approach is  the  same:   try to  relate  to the




explanatory variables the number of  days  of  each  type  of acute health




impairment due to respiratory disease during the fourteen day reference




period.    Results   are  presented for   separate   analyses  of  acute




respiratory disease  in  adults and children,  and a considerable amount




of  space  is   devoted   to  sensitivity analyses  of  some  of   the  more




important findings.



     Because of  our  interest  in chronic  as well  as acute morbidity, we




also investigate the associations between  certain chronic illnesses and




ozone, other air pollutants,  and additional possible determinants.  It




is  worth mentioning the measurement  of  chronic  health  status  as  a




dependent variable in the analysis below.

-------
                               3-5
     Unlike  the  measures  of  acute  morbidity  included  in  the  HIS,


information about chronic  morbidity does not lend  itself  so simply to


continuous measurement.   One  is  interested  not  so much  in "how many


days" of  cancer  or  heart disease an  individual  has during a specified


period of  time, but  rather  whether or  not  the individual  has  such a


chronic ailment.  Those surveyed in the HIS are asked whether they have


any  limitations  on  their work, recreation, or  other activities caused


by health impairments.   They are  also asked the nature of any ailments


reported.    Thus,  the  HIS  does   identify   individuals   with  asthma,


bronchitis, emphysema,  or  other respiratory  diseases,  as  well as with


chronic   diseases   of   the   cardiovascular,   nervous,   digestive,


musculoskeletal, or genito-urinary systems.


     Because  the data on  chronic  disease   are  of  the  "presence  or


absence"  type,  they  are most  appropriately  analyzed  using techniques


designed  for  qualitative  or  limited  dependent  variables.    These


techniques,  including  probit   and   logistic  analysis,  are  discussed


below.    With  respect  to  the  types  of  chronic  morbidity  to  be


considered, we  concentrate primarily on  possible  links between ozone,


other  air  pollutants,  and  chronic  respiratory  disease.    If  air


pollution  does  impair  health,  this  is  where  the  impairment  is  most


likely to  occur.   However, since some air  pollutants  have been linked

                         ii
to cardiovascular disease  , we  briefly explore  it  to see whether ozone


is associated  with  its  incidence.   One  problem we encounter there is


the absence of data on several potentially important confounding

-------
                               3-6
variables.  Finally, we briefly explore the relationship between annual




average pollution levels and several other forms of chronic deseases.




     The  Appendix   to  Chapter  3  describes   in   great   detail  the




definitions of  the  measures of health status  used  in our study.   It




also presents  the method used  to  create  these  variables  from the HIS




Public Use Data Tape.




     Tables  3-1  and  3-2  below  summarize  the  frequencies  of  our




alternative measures of  acute  and  chronic morbidity.   In  Table 3-1 we




compare the frequencies of morbidity in the  final  RFF sample of adults




with  the  corresponding frequencies  in  the  whole   sample   of  adults




receiving the smoking supplement.








3.2  Explanatory Variables



     Here  we   identify  and   discuss   the  sociceconomic,   pollution,




meteorological  and  other  variables  used  to  explain  variation  in the




measures  of  acute and  chronic illness.   Where possible  we  speculate




about   the   expected  signs  of  the  independent   variables  Ln  the




regressions below.




     Because of the  role they  may  play in explaining acute and chronic




morbidity, a  number of  individual socioeconomic  characteristics from




the  HIS  are  included  in  almost  every  regression  below.    These



variables—and,  in  parentheses, the  way  the  variable appears  In the




equations that follow—include:

-------
3-7
Table 3-1.
Summary of
Adults' Health Status
SMOKING SURVEY
Measure
TOTRAD2W
BEDDIS2W
WKSLLD2W
RADMINOR
TRADRSP
TBEDRSP
TWSLRSP
RADRSPPS
3EDRSPPS
WSLRSPPS
PRSCHRSP
PRSCHCRD

Measure
TOTRAD2W




























Quant i
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
Mean
.8508
.2921
.1866
• 3931
. 1741
.0775
.0500
.0426
.0257
.0227
.0750
- 1049

ty ?r















N
26,271
26,271
16,165
24,161
26,271
26,271
16, 165
26,271
25,271
16,165
26.271
26,271

equency
22657
734
639
403
228
172
77
201
36
18
87
14
20
15
970













Frequencies
Percent
86.243
2.794
2.432
1.534
0.868
0.655
0.293
0.765
0.137
0.069
0.331
0.053
0.076
0.057
3-692
Measures


FINAL RFF SAMPLE
Mean
.8818
.3129
.1952
.3969
.1674
.0137
.0557
.0414
.0251
.0240
.0692
.1071

Quantity
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
N
14,441
14,441
8,766
13,234
14,441
14,441
3.766
14,441
1U.U41
3,766
1U.W
14,441

Frequency
12415
384
366
220
139
102
35
110
23
11
53
5
11
6
556














Percent
85.971
2.659
2.534
1.523
0.963
0.706
0.242
0.762
0.159
0.075
0.402
0.035
0.076
0.042
3.350

-------
Table 3-1 (Continued).
                                        3-8
SMOKING SURVEY
Measure Quantity
BEDDIS2W 0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
WKSLLD2W n/a
0
1
2
3
4
5
5
7
3
9
10
1 1
12
13
14
RADMINOR n/a
0
1
2
3
4
5
6
7
8
9
10
1 1
12
13
Frequency
24559
503
359
201
112
98
31
75
30
12
36
H -n
i 3
1 1
9
221
10106
15237
395
192
54
3"
5 1
16
17
9
5
85
2
10
1
14
2110
22557
277
294
182
92
50
25
91
6
4
37
1
6
2
Percent
93-483
1.915
1.367
0.765
0.426
0.373
0.118
0.289
0.114
0.046
0.137
0.049
0.042
0.034
0 . 84 1
n/a
94.259
2.444
1.188
0.520
0.229
0.377
0,099
0.105
0.056
0.031
0.526
0.012
0.062
0.006
0.087
n/a
93.775
1.146
1.217
0.753
0.381
0.207
0.103
0.377
0 . 025
0.017
0.153
0.004
0.025
0.008
FINAL RFF
Quantity
0
1
2
3
4
5
6
7
3
9
10
11
12
13
14
n/a
0
1
2
•5
j
4
3
5
7
1
3
9
10
12
1*1


n/a
0
1
2
3
4
5
6
7
8
r*
7
10
12
13
14
S il-'.PLS
Frequency
13465
254
224
111
63
62
23
41
13
7
23
8
7
3
132
5675
8255
214
103
42
21
38
10
10
4
3
53
5
8


1207
12415
144
151
95
54
30
10
47
3
3
26
3
2
241

Percent
93.241
1.759
1.551
0.769
0.436
0.429
0.159
0.284
0.125
0 . 043
0.159
0.055
0.048
0.021
0.914
n/a
94. 171
'2.441
1.173
0.479
0.240
O.U33
0.114
0.114
0.046
0.034
0.605
0.057
0.091


n/a
93.811
1 .088
1.217
0.718
0 . 408
0 . 227
0.076
0.353
0.023
0.023
0-196
0.023
0.015
1 .821

-------
Table 3-1 (Continued).
                                       3-9
SMOKING SURVEY
Measure Quantity
TRADRSP 0
1
2
3
4
5
6
7
3
9
10
11
12
13
14
TBEDHSP 0
1
2
3
4
5
5
"7
I
3
9
10
12
13
14
TWSLRSP n/ a
0
1
2
3
a
5
6
7
8
10
14
Frequency
25153
328
262
141
85
61
18
62
9
2
16
1
4
2
127
25597
265
174
85
34 .
34
6
22
8
3
4
2
1
36
10106
15798
192
84
36
16
19
3
4
1
10
2
Percent
95.744
1 .249
0.997
0.537
0.324
0.232
0.069
0.236
0.034
0.008
0.061
0.004
0.015
0.008
0.483
97.434
1 .009
0.662
0.324
0.129
0.129
0.023
0.084
0.030
0.011
0.015
0.008
0.004
0.137
n/a
97.730
1.188
0.520
0.223
0.099
0.1 18
0.019
0.025
0.006
0.062
0.012
FINAL RFF
Quantity
0
1
2
3
4
5
6
7
8
9
10
11
12
14

0
1
2
3
4
5
6
7
8
9
12
14


n/a
0
1
2
3
4
5
6
7
8
10
14
SAMPLE
Frequency
13843
165
141
87
48
33
9
32
4
2
1 1
1
3
52

14078
128
102
52
20
22
5
12
4
2
1
15


5675
8556
104
48
22
10
12
2
4
1
6
1

Percent
95.859
1.143
0.976
0.602
0,332
0.229
0.062
0.222
0 . 028
0.014
0.076
0.007
0.021
0 . 429

97.486
0.886
0.706
0.360
0.138
0.152
0.035
0,083
0.028
0.014
0.007
0.104


n/a
97.504
1 .136
0.548
0.251
0.114
0.137
0.023
0.046
0.011
0.068
0.011

-------
3-10
Table 3-2.
Summary of Ghildrens
Health Status
Measures
FINAL RFF SAMPLE
Measure
TOTRAD2W
3EDDIS2W
WKSLLD2W
RADMINOR
TRADRSP
TBEDRSP
TWSLRS?
HADRSPPS
BEDRSPPS
WSLRSPPS
Measure
TOTHAD2W















•

•
•

•
•
•
•
•
Quantit
0
1
2
3
4
5
6
7
3
9
10
1 1
12
13
14
Mean N
4156 15.711
1913 15,711
2515 7,972
1159 14,283
2191 15,711
1200 15,711
1618 7,972
0707 15.711
0465 15,711
0664 7,972
y Frequency
13808
585
460
305
147
112
36
90
18
10
14
4
4
2
116











Percent
87.387
3.724
2.923
1 .941
0.936
0.713
0.229
0.573
0.115
0.064
0.089
0.025
0.025
0.013
0.738

-------
                                    3-11




Table 3-2 (Continued).
FINAL
Measure Quantity
BEDDIS2W 0
1
2
3
4
5
5
7
8
9
10
11
14
WKSLLD2W n/a
0
1
2
3
4
5
6
"
3
9
10
12
13
14
RADMINOR n/ a
0
1
2
3
4
5
6
7
8
9
10
1 1
14
RFF SAMPLE
Frequency
14578
429
310
171
74
43
11
50
10
6
5
1
23
7739
7130
362
225
97
46
54
8
^
0
2_
3
14
2
1
12
1428
13808
142
120
83
36
24
5
18
5
2
2
2
36

Percent
92.788
2.731
1 .973
1.088
0.471
0.274
0.070
0.318
0.064
0.038
0.032
0.006
0.146
n/a
89-438
4.541
2.822
1 .217
0.577
0.303
0.100
0.075
0.025
0.038
0.176
0.025
0.013
0.151
n/a
96.574
0.994
0.840
0.581
0.252
0.168
0.035
0.126
0.035
0.014
0.014
0.014
0.252

-------
Table 3-2 (Continued).
FINAL
Measure Quantity
TRADES? 0
1
2
3
4
5
6
T
3
9
10
11
12
13
14
TBEDRSP 0
1
2
3
4
5
6
7
8
9
10
11
14
TWSLRSP n/a
0
1
2
3
4
5
6
7
8
9
10
13
14
RFF SAMPLE
Frequency
14601
351
293
190
80
62
18
50
5
3
6
1
1
2
48
14980
276
206
117
44
27
3
27
6
2
4
1
13
7739
7443
217
158
62
24
41
6
2
2
1
7
1
8

Percent
92.935
2.234
1.865
1 .209
0.509
0.395
0.115
0.313
0.032
0.019
0.018
0.006
0.006
0.013
0.306
95.347
1.757
1 .311
0.745
0.280
0.172
0.051
0 . 1 72
0.018
0.013
0.015
0.006
0.083
n/a
93-364
2.722
1 .982
0.778
0.301
0.514
0.075
0.025
0.025
0.013
0.088
0.013
0.100

-------
                               3-13
      o  Age  (AGE)—Age  may play an important  role in explaining both




acute and  chronic  morbidity.  When  it is  entered  linearly,  one might




hypothesize that  the  older  the  individual  the  greater the  number of




days of  acute illness  (or  the  greater  the likelihood of  having some




acute  morbidity)   he   or she  will  have   during   a  two-week  period.




Similarly, older individuals are hypothesized to be more likely to have




a chronic  respiratory or other  illness  subject to aggravation  by air




pollution.  In both  cases,  the  rationale  is the same:  over time, an




individual's  stock  of "health capital" wears  down, making him  or her




more susceptible to acute or chronic morbidity.




      However,  one  need  not  confine  the  investigation  to  linear




functional  forms.   In some  cases,  one might  expect  a  quadratic (or




u-shaped) relationship between age and morbidity.   To explore for this




possible non-linear  relationship between  age  and  health, we  use both




age and its square of age (AGSSQ) in several regressions.




      o  Income  (INCOMCON)—Another  potentially important  influence on




acute and chronic morbidity  is income.  Wealthier individuals are able




to  purchase  higher  quality health  care  as well  as other  goods and




services likely to  contribute to good health,  including  better  foods,




more  leisure   time,  and  so on.    Thus,   we would  expect  a  negative




association between income and morbidity.




      The only information   sought by  HIS  from  the  respondents  is  an




indication  of which of eleven intervals their  annual household  income




falls  into  (see Appendix  A at  the  end of this report for  details).




Rather  than  work  with  ten  dummy  variables  in  each  regression,  we

-------
                               3-14
converted the interval  data  into a continuous measure  of  income using




the midpoints of  the  income  ranges and assigning  an income of $30,000




to  everyone  in  the  "$25,000  and  greater"  category.    It  is  this




"continuous" measure of income that is used in the regressions.



      o  Education  (SDCOMCON)—Education may affect  health a number of




ways:  by familiarizing individuals  with the early  signs  and symptoms




of  disease,  by  familiarizing them  with  preventive measures,  and  by




impressing  upon  them  the  importance  of   periodic  medical  checkups.




Also, education might also help  people  recover more  quickly and easily




from diseases they  do contract—make  them  more efficient  "producers of




health,"  in other words.  Therefore,  years  of education completed ought



to  be  negatively associated  with  acute  and chronic morbidity.   (This




was also  converted to a continuous measure  using  interval  data from the




HIS.  Again, see the questionnaire in the Appendix to Chapter 2.)




      o  Race (RACSW1B0)—Although it is not clear which direction such




a relation would take, race could play a significant  role  in explaining




variations in health  status.  RACEW1B0  is a dummy  variable taking on



the value unity if the respondent is  white, zero  if black.




      o    Sex  (SEXM1F0)—Differences  between   the  sexes  might  also



account for variations  in  acute  and  chronic morbidity.  To account for



this possibility, the dummy variable SEXM1F0 is included which takes on




the value unity if the individual is  a male, zero if  female.




      o  Marital  status  (MARY1N0)—The  presence  of a spouse may enable




one  to  recover  more  quickly  or easily  from  an  illness,  and  some




evidence   suggests  that  married  individuals  may  be  healthier  for

-------
                               3-15
psychological reasons  as well.   To  test  for these  effects  the dummy



variable MARY1N0 takes on the value unity if the respondent is married,




zero otherwise.




      o   General  condition (FAT)—As  discussed earlier,  it  would  be




desirable  to  have information  about  individuals'  exercise habits and




physical  condition.    The HIS  provides no  such information  but  does




include  the  height and  weight  of  all  adults.   As  a  crude  proxy for




physical condition, we  have created the variable FAT,  equal  to weight



in pounds  divided  by  height in inches.   Since overly thin individuals




may,  along with  the  obese, be  prone to  illness,  we  generally use a




quadratic form including both FAT and its square, FATSQ.




      o   Smoking  (SMOKY1N0, NCIGSDYN)—No  environmental influence has




been more  clearly linked to acute  and  chronic morbidity than personal




smoking.   There  is  even evidence  that  proximity  to   smokers  nay  be




injurious to health.  Accordingly we measure smoking  in  one of two  ways




in  the  analyses  below.   SMOKY1N0  is a dummy variable  taking  on the




value  unity  if   the   respondent  currently  smokes  occasionally   or




regularly, zero  if not.  We  also measure smoking  on the basis  of the




number of  cigarettes  an individual usually  smokes  each day,  NCIGSDYN.




A Prj-ori we expect a positive association bet-ween smoking and acute and




chronic illness.



      o  Chronic illness  (CHRLMDUM)—When analyzing the  determinants  of




acute morbidity   (work  days lost,  bed disability  days, etc.),  it  is




important to know whether an individual also has any  chronic illnesses.



If so, we  would  expect  him  or  her to have more days  of  acute morbidity

-------
                               3-16
during a specified period than  if  such  a condition(s) was missing.  To




account  for this  possibility,  we  include  in our  analysis  of  acute




morbidity the dummy variable CHRLMDUM which takes on the value unity if




the individual reports having a chronic illness, zero if otherwise.  In




our analysis of chronic morbidity,  a variable like this one becomes the




dependent variable and hence is not included among the regressorg.




      Several  other   individual   or   household   characteristics  are




available from the HIS.  Although of less potential interest than those




already considered, they do appear in some of the reported regressions.




They include:




      o   Veteran status  (VETY1N0)—A  dummy variable  taking  the  value




unity  if  the respondent  is  a  veteran  of  the armed  services,  zero if




not.




      o  Central city residence (CITYI100)—To see whether residents of




central cities are more or less healthy than others, we include in some




regressions a dummy  variable  taking the value unity  if  the respondent




lives in a central city, zero if not.



      o    Paid   sick  leave  (INDOCCAV)—As  discussed previously,  the




amount  of  paid  sick  leave  to which  an individual  is  entitled  may




influence  the  amount  of reported  work  loss due  to  illness.   INDOCCAV




measures  the  average  number  of days of  annual  paid sick leave in  1974-



received  by  someone  who worked in the  same  profession and industry as




the respondent.  This data comes from a supplement to the  197^ HIS  (see




Chapter 2 for the precise reference).

-------
                               3-17
      o   Profession  (BLUE)—Individuals  working in certain occupations




may be  exposed  to more  accident  and health risks  than  those in other




occupations, and thus experience more acute and/or chronic illness.  To




test this  hypothesis,  we include in some  regressions  a  dummy variable




BLUE  taking  the  value   unity  if  the  respondent  is  employed as  a




craftsman or kindred worker, equipment operator, farm or other laborer,




or service worker, and zero if  employed  as a professional technical or




kindred worker,  a manager  or  administrator, salesperson,  clerical or




kindred worker.




      o   Household density (CROWDING)—The  closer  the contact between



individuals, the more likely it  is  that  a communicable disease will be




transmitted  between  them.  To  allow for  this  possibility,  we created




the variable CROWDING  by dividing  the  number of persons  in a household




by the  number  of rooms in the  dwelling.   Our expectation is that this




measure  should   vary  directly  with acute  and  chronic  morbidity (the




lower the density, the less  likely the incidence of morbidity).






      o    Ozone  and   other  air  pollutants  (see  Chapter   2 and  its



appendix)—The measures  of air  pollution  used in  our analysis of acute




and chronic  morbidity  are defined  and  described in considerable detail




in  Chapter  2  and  its  Appendix.   Because of  the  large   number  of




different  pollution  measures  we  employ,  they are  not  discussed here.




Headers  are  encouraged  to  familiarize  themselves  with  the  pollution




measures before preceding with the analyses.  All air pollution

-------
                               3-18
variables  may be  expected to  vary  directly  with  acute and  chronic




morbidity.






      Measures of  meteorological  conditions appear  in  virtually every




regression.  Among the most important of these are the following:




      o    Temperature  (AVMAXTMP)—Temperature,  particularly  extreme




temperatures,  can  influence  acute  morbidity.   Accordingly,  several




different  measures  of  temperature  are  employed  in  our  analyses.




AVMAXTMP  is  the  average of the  highest daily temperatures  in  degrees




Fahrenheit  during the  two-week  reference  period for  which data  are




available on acute morbidity.




      The  relationship  between  temperature  and  acute  morbidity is  a




complex one.   During the summer months, for  example,  one might expect




hotter temperatures  to  be  associated with more acute  illness.   During




the  winter  months,  however,   warmer temperatures  might  signify  more




temperate  climates  where illness might  be  expected to  be lower.   For




this  reason,  we  created two interactive/dummy variables,  TMAXTEMP  and




DMINTSMP.    The  former  measures  maximum  temperature for  individuals




interviewed  in the second  and third  quarters  of  1979 (April-October).




The latter is the operative temperature measure during quarters  1 and 4




and  measures the  average  minimum  daily temperature.    This  approach




allows one  to test whether high temperatures  are  the  relevant  measure



in summer months, while low temperatures influence acute morbidity in

-------
                               3-19
winter months.  Appropriate annual measures are used in our analysis of

chronic morbidity.

      o  Precipitation (AVPRECIP)—Snow or rainfall might also increase

the likelihood or  amount  of acute and  chronic  morbidity,  or influence

work  or  school attendance  even when an  illness  is present.   To test

this possibility for acute illness, we include a measure of the average

daily amount  of  precipitation in inches  during the two-week reference

period.     As  in  the   case  of  temperature,   annual   measures  of

precipitation are included in the analysis of chronic morbidity.

      o    Humidity  (HUMIDRF)—This  variable  measures  the  relative

humidity  during  the  two-week  reference  period.   As  in  the  case  of

temperature,,  annual  measures  of  precipitation  are  included in the
                                              »
analysis of chronic morbidity.

      Several  other  kinds of variables receiving  mention  in Chapter 2

are used in our analysis.  They include:

      o   Indoor  air pollution (PGASPIPD,  PGASALL)—Indoor exposures to

certain air  pollutants  can be high, and  gas  stoves are reported  to be

responsible for  high  indoor levels of  several  of  these pollutants,  in

particular    nitrogen    dioxide.        Therefore,    PGASPIPD    and

PGASALL—respectively, the  percentage of  all  occupied housing units in

the individual's SMSA that use piped natural gas as a cooking fuel, and

the percentage cooking  with piped as well  as  bottled natural gas—are

sometimes included  as independent  variables in our analysis of chronic

morbidity.   A priori,  the higher this  percentage,  the more likely the

presence of a chronic respiratory disease.

-------
                               3-20
      o  Air quality information  (PSIY1N0)—A  dummy variable taking on




the  value  unity  if the  PSI  (Pollutant  Standard  Index)  is  reported




regularly in an  individual's SMSA,  zero  if not.   As  discussed above,




the expected sign of this variable is ambiguous.




      o  Availability  of  medical  care (PCDOCN)—This variable measures




the  per capita  number  of  doctors  in an  individual's  SMSA,  and  is




intended to  serve as  a  rough measure  of the  availability of  ;nedical




care.




      o    Airborne  allergens  (POLLENHF,  POLLENAN)—Both  acute  and




chronic morbidity  may  be  influenced  by natural substances  in  ambient




air to which people are allergic.   To test  this hypothesis, we include




the  variables  POLLSNRF and  POLLSNAN  which measure,  respectively,  the




ragweed  concentration  during  the  two-week reference  period and  the




annual average pollen concentration.




      The variables discussed  above are  listed in Table 3-3.   Beside




each variable  we  indicate its expected sign in regressions explaining



acute and/or chronic morbidity.



      Tables 3-^ and 3-5  below present  the  mean and extreme values for




these  independent  variables for  both the adults'  and  children's  data




sets.    They   indicate   substantial   variation   in  the  independent




variables.  For instance,  the  mean  value  for  03NB01, the average daily



maximum  one-hour  ozone   concentration  during  the  two-week  reference




period, was 0.0^5 parts-per-million—well below the current ozone

-------
                               3-21
Table 3-3.  Independent Variables
Expected Sign
      Age (AGE)


      Income (INCOMCON)
                                        s

      Education (EDCOMCON)


      Race (RACEW1B0)


      Sex (SEXM1F0)


      Marital status (MARY1N0)


      Physical condition (FAT)


      Smoking (SMOKY1N0, NCIGSDYN)


      Chronic illness (CHRLMDUM)


      Veteran status (VETY1N0)


      Central city dweller (CITYI1C0)


      Paid sick leave (INDOCCAV)


      Profession (BLUE)


      Household density (CROWDING)


      Air pollutants (see chapter 2)


      Temperature  (AVMAXTMP)


      Precipitation  (AVPRECI?)


      Humidity (HUMIDRF)


      Indoor air pollution (PGASPTPD, PGASALL)


      Air quality  information (PSIY1N0)


      Availability of medical care  (PCDOCN)


      Airborne allergens (POLLENRF, POLLENAN)

-------
                                                3-22
Table  3-4.  Mean and Extreme Values for Independent  Variables:   Adults Data  Set
                                MfTAN
                                        n n£V
03MR04




SPN"ni
S4MC n]





S4NP03





54NOII4
IJ3NPI1 1
AVM *
   •TNT m



   ix rci'p
           1 1^77
           1 07S«




           ! f 7 q P




           I n T R B
1 -j ° * ?




1 i' ? 6 ?




1 f1
           1 4 1 =15
]





1 <* 1 ^ ?





1 4?04
                         47. 1 "5*.4n J 07
                                  17
                          n.
                                               1 .4(1141 104
                                              17,4]
                                               0.
                                            0514, 7BQ
                                                                 4.J>1
                                                    • 1 3,n7142«57



                                                     n.'
                                                                                     n, 43000000



                                                                                     i.i:
                                                                                      . 1 3Annnnn
                                                                                   "21 ,
                                                                                    1 1 ,4ni*ftftnn
                                                                                     , nnnnnnnn
                                                                                   1 fl, nnonnnon



                                                                                30000.nononnnn

-------
                                           3-23
Table 3-4.  (continued)
v^»i/vpi_F n MEAN STO oev
SFX^'iro 14441 U«46527?4Q 0«49flpf>991
4f?E 14441 4? 1('7«.l 09"<9 1 * i 33753(1 1 2
A3£S<- 14441 21 S3 • 1 R77293H ) 740 < ?49?n 1 5fl <
FM I4.os J.annoiBO', n.43n,?i*A
r»T^n 1,A,, S*7«<07445 2.1*7Wflft
M4Ryi»iQ 1444] n,ft]ns9953 0«4fl7^fc30fl
* P T Y 1 ' 1 n Jo ]74 ^,Jflfl«,B^JR 0,3'124P3^
S"0"viMn 11*47 fi,i*«7^]rj ^ ,479(f>51 **
1%IC 1 ('•SOY"' lj?l] 6« '60?'"Sfl4 l?t1?0'7S34
Cu»j| KPIJU 1444] C!«!7'CI"'12? C(3"3''#>9J3
CITYJIO'' 14441 n,44T7Tft«;fl n,49Afl4l'iS
^ 5 T Y t f 1 1 j 11"1?! fl«fl4^Q?f1fl ^,3^t^^^*^
•^*_U? QO^iA 0«4*»1J736J' OjA^flSlTT^
t, ^ 004(1174.

n i.nnnonnon
I7.nnononoo 99.oo«onoflo
ino.noonnnnn 9fll" .nnnnnnnn
l.?«0««07 s.Q91«-««73
,.^,??M] 3t.?M*««7
rj i .nnnnonon
™ • - v
n t.oonnnnnn
„ qfl.nonnnnnn
, l.^OOOno"
ft 1 .nnnnnnnn
„ T.nnnnnnon
, I.«0«0nnn0
n 4flnonnnn n.innnnnnfi
p r Q " f t
                                                            . n m 1 1 1
                                                           1.0^06451"
                                                                                    nn

-------
                                               3-24
Table 3-5.   Mean and Extreme Values of  Independent Variables:   Children's  Data  Set
iiju^twB
S E i "•'1 • '
AG?
            12<>b3

            1 !

            1'

            1 i
            flan?
S"WM
3 P M e r 3
5 p M R p a
S a M y«i
S4NS,-3
s u N. 3 n u
* 2 -•'« " '
? 2 ^ » n 3
S2""5"a
J.:3f.'pn 1
\; ?M w n 3
r a \' s c t      1 1 7 u 5
70\=nj      ll'a?
            1 1595
            1' 5 * 5
            '53^5
           1571:
           1571 1
CI*v'ir..i   157! 1
°OLL?'
                              0 , 37<
                              0,0?107h58
 7 1 . I 7 •)
 K.-57ftOi;5!7
 1 2.5(5e73960
! "o.25l3!11s
 2o, 1 aa2»23
                                                     tt.
                                                                      o.isa
                                                                      O.UJO
                                                                      u , I U e
                                                                    2oO,09^
           1571 1
                              /,ori20037f

-------
3-25
                                              •vj
                                               • '
                                              3
                                              a

                                              o
                                                  3.


                                                  X

                                              O
                                               •
                                              o
o
C2
X

-------
                               3-26
standard  of  0.12  ppm.   For  a number  of  individuals,  however,  the




average  daily  one-hour maximum  during the  reference period  was  zero




ppm, while for  at  least one individual 03NH01  took  on the  value 0.251




ppm.  Figure 3-1  presents  the frequency distribution  for 03NB01.   The




highest  single  hourly ozone  reading  experienced by  any  individual  in




our final sample was  0.43 ppm (indicated  by  03NR03,  which measures the




highest reading during the two-week reference  period).   Although we do




not discuss them here, considerable variation  also exists  in the other




pollution  measures,   the  meteorological  variables,   the  socioeconomic




characteristics  of the  population,  and  the  other  variables  we  have




created.




     A final remark about the  independent  variables  used in this study
                                       4



concerns  the  correlations  between them.   When  close (but  not exact)




relationships  exist  between two or  more  independent  variables,  it  is




possible  to obtain least-squares estimates of  their  parameters but the




estimates  will  be very  difficult  to  interpret.    This  problem  of




multicollinearity,  as  it  is  called,  often arises  in air  pollution




epidemiology because  concentrations  of  certain pollutants  often  move




together.   (This  is  due in part to  the fact that  they often originate




from the  same sources, e.g., automobile exhaust, power plant emissions,




and so forth.)  Before discussing the statistical analysis conducted in




this  study,  we   present  in  Tables  3-6  and  3-7  respectively  the




correlation coefficients between all  pairs of independent variables for




all  the  adults  and   children for  whom  data  are   available on  two




particular variables.   (The matrix of  correlation  coefficients for the

-------
                                                                                              3-27
                                tr
                 c  c  4:    ^  c -c
                 —  <:  —    <  c —
                     c
                          x    a
                          lu    Z
                          U)    CV
                          I    (/>
                          c

                          u.
                          a    •«
                                c
                          2    3
                          O
                                O
                         o     a

                         c     -*
                         X     v
                         Z
                 p- — •»

                 X C  p-
                 3> o (J.

                  • o
                 c
                 ai — IT
                 p- C P-
                 tf c ^
                 — e •<;

                  • c
                 o
                 —  c

                  •  c
 •£  —  -3
 t~  C.  -O
 tv  c p-
 
c
 I

p-  — {ft
P-  O »-
a  c if
(VI  C <
—  — •*    *»  p» C

•c  c p-    4-  ru »>
p'  c o    e  •£ c
p^   .       c   •-.
 •  c         •  c
c           c
 I
c  c r-    (v  •c p-
-»  c o«    e  p- c
~   •       c   • —

c           c
 I

»  — *    P-  — C

•—  C P-    C  C P-
C.  C O-    4  C <5
(V*   •       C   • ^«
 •  o        •  o
c           c
 I


•A  C p-    C  C —
^  C ^0    C  C IT'
^  C *    P-  C. P-
o   •       -•   •
 •  c        •  c
<=           o
tflCP*    «OP»    XO  —
fcin    ectf     *sxi
tnc^    ^o^    tpcp*
^-»   «        o   •        >^    •
 •  o         •  a         • s
c            o            c
                                          P-CP-    P-OP-    -tCP-    O  C —•
                                          3} C>  tl^    9 O 1^    C^  O l^    O  O tf)


                                           • o         *o        *o        *o
                                          -* —   — e    ic  ••« «-    (vdO
P-OCVJ    ooro    c?2 —
CP  G —•    CO^-    CC —

 •  c        •  o         •  o
o           —           o
                              cc*    -co-*-    o«>
                              crcrw    p»C5(\i    if  c -~
                              CO"    »O-«    IS —
 O
CJ
                CO—    .-VI —  -«    >.-,,.    -»-«T
                CO*    ffO*    O^-»    P"O^
                OO(M    •4JC'V    XCOJ    ^  ^ —*

                C~"«—    ?    .  —.    J  ~  —.    (VI   •-.
                                                                                 a
                                                                                 
-------
                                                 3-28
 I


 01
—t

"5














~
c

^
^
.X

1/5
Z
C

u.
c

z

X
^
—
2
^
*J
~
•22
a.
o
^
^
c
^
—

J
_
t
A
i .3
0
a
t x

v)
»_
—
! ^
U
U.
u.
u>
c
(J

z
-
<

j ^J
3^
^
O
















I
c
a
(VI
^
c
2
IV)



tn
c
5

CM
U)


_
^
a
4_
CM
(/>



^
C
2;
^*
^
;/)


O
Z

VI

—
II

4
yj


j.
Q
a
i
CL
y»


c:
~
^
0.



o
I
^
a.
s/i


^
C
^
."1
c


c
a
i**
c


^
c
;v
~
."^
£






C
f-
c
i^

c
-*
o

e

«*
t~
(*4
fH
C
«
c

z
X
•o
*c

•
—


o*
!*•
«•-»
ir

»
c

o
£
«""
•
e
*
^
X
.-»
•
o

r*
JQ
r*-
—
**.»
o

o
o
c
c.
•
—

r*
iv
^4

nj
Ml
•
C

X
(Vi
«
c-

T
v»
*»-
*-
rw
•
o

^
_^
^
^
—. tf.
G (^
G IT
C. ^
C
— c
a A.
C: C
« ^-
G


C C
c. 5
s »-
c. c
• .-•
—


~ c
c; i£
C f*
c o
* «—
c



*- c
c —
c x
C f-
«
c


o •—
c u"
*
c

.— 0
C lA
0 r*~
•
c


— (M
C- P--
c-. r-
c <•*
c


O f-
C f*
0 I*"
* —
1C


«••» 
O — •
O —
• .—
C






r» .- u*.
.- c r*
•* c tr
* c
o
(V. * G
4 >C X
C •£ C
O B •-
• o
c

0  — 0
o e K
c c p-
4 C C
o • «^
• c
c


IP — 0

O C X-
•«; o P-
*•• •
• c
«

^ ^™ C?
ec o —
"*5 C" X-
if. o p-
••• B
• o
c
XI — 0
O 0 'ft
•; c r-
*M •
• O
^

C 0 (Vi
-TCP*
<^ G P*
C C <•>
• o
—

X. c- P-
P- C P*
•— o 1— 1
^ • **
• c
e

O> C P-
T C P-
o> e r>
c « —
• 3
O

vT — S
0=2.
C C ^«
M • -•
• C
e

-C c: —
30 —
(VI • —
• C
c:

•»• — • X
I/ C- ^
T C —
(^ C -*
A. • —
• —
^

-J-
O
^
2
(Vt Z CVi
(V. -» r*
c •
• c
c
c — e
 •
• c
c
c <= *
0 C Ml
C. C »-
o •
• c
—

-r. — c
cr c ^«
O C tfl
•c c »-
• •=
=

P- C. ^»
— <= rf>
X. 0 t-
'^ •
• o
0

c o —
cox
p- e. r-
f-» 9
• O
c

o- —x.
3- C X.
a- o sO
O •
• O
o

aj o •«
•** *
* c
o

yf *•* ^^
^ C. f*-
T O ^*
*" O 
^ c ^
!>*• O f*
(Vi C ^>
M *
• G
C

<*• — c
— c <»
sr c n
< c >c
.-* •
» o
G


IT *- -*
< G ev
O C- 
c c »*
o *
• G
C

0 C (V»
c c *-
c •
• o
~"
* s^;
o o «n
c o P-
* •
« o
0

p- — c
sc o —
£, C :f)
*i o r-
• o
o

if 0 —
srv c x*
p- C P-
ff^ •
• 0
o

O
a •
• c
c.

<»• e x
X. o *
*^ •
• 0
0

•» — X.
»- e p-
— o /*
— o •&
^* B
• O
—

» (
—
I
z
(M
a.
0
o
c
a.
IT
IV]
•
c

f
c
o
pb
—
B
o

„.
^
f-
ru
ru
«
c


c
c
o
c
G
•
—

i
<_;
a-
•
c
c
c.
o
o
•
M

^,
^
J
"*
"^
o

r,.
«—
X.
^»
•
0

s
c
p-
^™
«
0

J
~
^
o
•
o

CM
M
•
O

X
/fc
3";
»
*^
•
3

_,.
o
i
2
X (VI
c
— c
c •»
C .
C -T
c <-
o ^
•
o


— c
C -T

c i
•
c



c -
O P-
•
o


— X.
0 X.
e -c
•
o


c -f
o c
9
C


— X.

• IJP
c ^
«
<^






c
o
e
i
p-
•c
n
X.
•
c

a
p-
PI
tr
A
•
o

c
c
c
c
o
•
•M


^

c
s
^
*.
o

£
•0
^
O
•
o

(**
c
«
o
•


^
M
c
•Xl
•
^

*
•M
•
o
i
•w
O
x
tj
.H
•
C
t
_
^
^
™
C P-
C
— I
=• I
e e
• r*
C


— c
o a>
c r-
o c
• —
e


c 
c p-
o c
• f~
o



— o>
c •»
O P?
G O
•
G


G -*
G ,0
•
G

~"-
o m
0 «
•
a


— o
C 1C
C P-
0 O
0


e 
P" C. •£
*• B
B O
0

-* O -»
IV O -O
•M B
• 0
c
« -e
O 0 <•>
p» O •»>
CM •
• o
o

 c o
u^ » •••
B O
o

p- — C
(VI »
B O
O

CC O *»
c = «
(V) B
B O
=
c -0
P- O (*>
X 0 O
IVI B
B 0
0


^V B
B O
^
1
f
—
3;
X
C c; X
o c X"
c: c tr.
C B
-»•£<*
r" T CT
IV c P*
O B
B 0
o
I
•£•»•*
1" c C-
ff e p-
» 0 •••
O B
B O
O
1
-* — -*
~ e •>
c c p-
If 0 p-
O B
B 0
C
t

cv s ru
cv •» «
— f.
C5 B
B e
o

^ o *•*
•r o tf>
^* B
B O
°
cv. T nj
•^ ivt CV
(V! —X.
O B
• 0
o

P** — X.
••* v .***
^ G X
* * *
, 3
c

••SI"
P^ C X*
(f C (Z
f* B
B O
o

e! e" (•»
4 C tf*i
O O 3)
m B
• o


T — I
"O G •£
vT G t
— * •
* G
C
— ~ X
>C G —
^» C X
^ 4»
• O
o

^- — X
c c »;
— G *-

< e p-
X. c P»
O B
• C
C
1
c ac ~r
cy cvi o
n c p-
(T C P»
O B
B O
c
1

< — CM
•£ X. P1^
^ X< IV*
•~ ** X.
o «
B C
o

* o£;
•c c 4>
*^ B
B S
c
^ -«
 in <\«
— — if.
0 B
B O
3

•£ — X.
(•» * ^*
G — t/'
IT G (»
* cr
G
cr *•• if i

o c u1.
X G (t?
(*T •
• O
O
—
-r G f*1*
G 0 rf*
rfl — ffl
(*> •
* G1


P. - 1
s c —
~m O CO'
•M B
B O
o
cr — 
0 B
• O
o
1
•* cc -r
» •* »
•C f- P-
— — p-
0 B
B O
o
1

o P- ru
x. c i"
•c o cv
«» C If!
C B
B O
O

^ C C*1
P* C IV
P~- C rf>
•M B
B O
o
a> p- cv
•C 0 CVI
•* o &
o •
• c
o

tf.
•^* O C"
^ G 10
>.*- O  0 X>
CVI B
B C
c

« — J3
— c -i
^ -c •-»
X 0 T
«^« •
B O
^

_^
^
^
.i
P" — p-
« e cc
X C c-
«v; c o
O B
B O
c
1
•» — X
ty c •»
X e «
x. o c
C B
B C
a
I
tf. tt X
P- — «T
*-« O «C
fr c :>
O B
• O
o
t

•c — -a-
« c rv
e e *
P» O -^
O B
B O
O
1
a. *o <*
C* G *C
•- — «c
G •
• G
G
< -..»
= c -o
*- c. •&
e B
« 0
o
1
f" — -I-
G 0 C-
— — e
— c =
B C "~*
o

I 0 «*l
3- 0 (?
IV) O C.
M B —
B O
0

O O IB»
— CO
— c c
CVI B —
B 0
c

nj -. ^
c o o
sec
O B «•-•
B O
o
1
P» <£ 0
c .c o
C B «.
B O
o

•C  C 0
B O
o

CV) e c"
P» 0 »
O C 0
M * —
B O
o

P"l C l»>
Xi O ^
•c o o
V4 B f*
B c:


•V) — 0
o o o
"1 O B-
(*» C O
^B B #^
B C
o
1
P*> B- <=
10 =
* o s
O B —•
B C
c:
1
XV — 0
p- <= c
r»* T- — •
X' C C.
™ » *^
B O
O
1
^
/^
^
c

-------
3-29















































^_^
-3
«
^
••H
ij
O
o




o
i

„,
w

X
U.
C
z
^
^
~
V.

;-;
*
X
«•
o
—
c
r~
^

J
—
A
T
~
J:
**
x

• tf
*—
cr
jj
^
u.
Jj
3
U

-r
C
*—
»—
4
—
I
i
5






\


]

'
•
i


c
CL
Si
IV
(ft
IV
(V

c
^
U1


X
~
1^1



I

'J»


^
c
5
5;


,-,
c
~
^
2_
1/5


~
^
a.
,yi



-*•
c
^;

;


^
^
-


=
.7
•"
C







f\<
in
m
c
c.
c
1
*T
ni
i
i
c-
e
t
E
=
•
c
1
IM
?
•
c


7
j
*
^
•
c
&•
c
f\J
c

if.
r*-
c
^
IM
*
c

|
v
rvj
•
*


C
.i
"•-
r*1
*
C
1
—
r*.
^
c

?
,/
c
•
~



*
2
-
C CO
C. (V
c (C
•
o
•c o
c
— if!
c -*
c -c
e 0
*
c
C -4-


c 
c o
C


•— •*•
O (*>
o *?•
c c
• —»
c


o r*
c o
• —
o



•— i
c. c:
C- •-•
=, c
c


O C;
0 c:
^


•= =
— c
• «•

0 C m
e rr P-
» o
o

<* ~ o
IV* O I*1
•C o .p"-
s» C «**
r— • •-•
- 0
o

X o ***
0* 0 ^
0 C «•>
<\; • •«
• c
o


^± •— .r-
c c ^
-* O OJ
O C •—
• c
c

!•» C -»
3- c -
Jj
z
tv* •— if
•<[ C IT
o e in
c c to
• c
o
h- 0 *
p- c. P-
p* o o
• c,
c
— CX
x c f-
ar. e o
• o
i
» — x
e c aj
•0=0
• o
c
1
•c c —
X O Jl
c •
• e
0
in — • «
(V O —
^ c ti^
e •
• 0
a


•* » X
ir o f-
• e
^

» c ru
•* 0 —
• 0
<3

^- c ^**
ir o rv*
•""s ^- ^™
if* • *•*
• 0
^

^
^
*-
.m
r
X
e*
o
I
A.
e
r-,
c
i
x
c
i
i"
„ ,
•
0

•c
—
•>
o


o-
C1
•)
o

•J.
o
O"
w
o

ir
p*.
rv«
^
c
•
o

»
C\4
•^
«
o


*f
ftj
^~
?
.
e

IT
n;
^i
o

J
.—
J*
•
—

a
^
a.
^
^
O tft
O
•- X
C X'
e i>-
<= o
c
C 05
C O
c.
0 X
C 0

e •—
e x
.
G

•»« X
C If!
«
-s



o —
O lf>
•
o


— o-
C: fl
c **
C !*)
• ft
o


o «
e p-
C' (*)
• •««
o



-^ *™
0 -»
O IM
C ^*
o


c -»
O <\l
o •—
c


c •»
e — "
V •-•
o







x •-• in
r" o in
(V c x
•»• •
• e
c
C •— X
_ P- S
C X O
c • —
• c
e
cox
<•• ~c x
— CV f
(\J C C
7
e — x
(V P** X
C •£ O
O • —
• o
1
•c o —
-4 O Ifl
IM »
« O
O
1
•« o —
»- c. tn
ou •
• o
o
t

IV — X
*£ C *-
f\J •
• C
o
t

.» c m
-cot*
c • -^
• c
c
1
X XI >
IV C« r7;
•^ X P1-
o x m
c • ••*
• o
e

-C 0 P~
4 O (*)
o • •••
• o
o
1

t— ~ —
1^ O •»
con*
^» o •—
• 0
i
1
-.<=,*
••* O f\i
f. 0 —
• e
o
1
(•• c *
r c IM
<*•£;*-
••• • — •
• o
O
t
a.
~
u
—
•-»
••(MC
•<» if. a
If . If: -9
~- — X
O •
• o
c
* o -•
c e c-
(MOO
• c
•* — f.
e c —
u: o r-
If O C
O ••—)
• c.
c
o o *~*
r- e. c
• o
c

a1 o -c
a e •*
—. •
• o
o


IM •
« o
o
I
^ — X
if c  c o
c « —
• e
c
I
c p* o
(\i e e
o • —
• c
c
1
r" X p-
m e o
• o
*
x — •»
0 •
• o
o
1

o •
« o
o
1


o •
• o
^
1
P- — CJ
X 0 -T
^. o ;n
IT 0 i"
G • -*
0
1
•* -" IM
f- O -*
c o in
X ci m

• o
c
1
c c -»
c c in
•c c .->
c • —
t> o
0
1

ifi ^ O
<•* ^ tf*
»  C 3>
C •
• o
o
i
c ^o
o •
• o
c
1

— •* r,
fr x i»»
o •
• o
^
1
1" »- *
x = m
IT 0 m
'/i O (U
o • —•
• c
c
1
.- « ^;
— e in
-* c m
IT- 0 <\i
C • —
• O
0
1
e o tn
•com
J"* C (M
<= « —
• Q
o
1

o ^ c.
>*• X f*
O P» 0
— (M C
• ^
o

X O c
_ 0 r~

-» e o
• c
c
1
x — «e
— rw c
o • •-•
• o
c
f cy a.
O Ai O
X O lit"
tv. c o
c • — •
« o
e
*
<£ e m
c •
• c
c
1
p- c r"
0 •
• o
o
1

»- c «•'
o •
• c
 ••^
0 C •«
If! O CVi
O J^ •"•
• 0
e
1
f if *
m (v (v.

• e
c
t
e p- -*
f^ -j- ^j
c. p* -*
~ • **
V C
—
1




^J
0 X. ifl
a if . if

-------
                                                                                      3-30
       c
       x
       (M
       WJ
tc.  f-  n    it, •£  i"
*  c  in    F-  x  if
i"  i"  «*•    if  «-• in
  x  ir    (\i i  if.
—  —  c    — — o
c    • —    c   • —
                            r>  ^  Ifi    P»if?G     P*  JVi P-     (V>Xr    X  »•• X;     *  — C
                            •4  C  IT    —  < C     X  
o
X.
3


A


3
(•»
c
I

(V
t/1


=
•Z
s.
t/1


c
a
IT
^
y>


^
o
5
^
in


^
^
*»•
•^i


^»
=
~
x
VI


(*>
c
I
^
3L
^
vC
nj
c
c
•
c

*
c
C.'
*
e

(M
m
•w
O
•
O

r-.
r»
^^
c
•
c


S
<=
•
O
1
c
IT
o
•
c
1
•c
t?
(\!
c
•
o
<
.£
c
G
(V.
•

O P*
•
O


P- *
C —
s s
• —
c


p- *
o —
— ifi
c (••
• V*
IT
rv.
f"
C1"
c
•
c

<-
t"
c
»
c

X'
<«5
^«
O
•
a

7
0
c
o
•
o
1
X'
(*>
,»•
o
•
G

x.
CVI
o
c
•
c

f
(V)
•"
^v
c
X P-
C 
•
o
•
0

ir.
*•
c
•
o

c
nj
r*^
O
*
O

X
r*
•3
n
o
~. x
^» l£
^ F*-
^ C
• ••*
C


a x
X; O
« «l-l
o


•£ (V>
C if

r>
c
*
s

c
(\.
o
•
o

-*
•«•
\f'
o
c:
•
c
I
o
ru
*
•
0

o
*•
c
c
•
c

^
u"
*
£
O
v» £
^- IE
C *-
C C
• 1—
o


C X
0 C
• r*
o


c a»
X. IT
<= f-
*
C


<= *
^- ru
^ |C
•
o


s *>
1J d"ji
C ^>
•
0


— f\J
^ p*.
0 "n
• r«
C


— (^
c r^
3 f*-
G 1*1
• *•*
^*
C
^C
lf>
c
•
c
1
^
c
c
•
o
I
c
f\J
£
c
«
o

<7
-c
rvj
^^
•
o

PU
CC
c
•
o

<;
f
w
cr
*
o

(V
G
T*
rv.
v
•M Xfi
C (^
C 4
C O1
*
c


* *-.
1^ J
«
^5


O t*
C -G
C ^J
•
c


—     X.  •£  X>
Li~.  — -T    C  C  P»
—  (V (•*    —  •«"   If.  —
                                                                                                                  X  (VI  C
                                                                                                                  If  T  X
                                                                                                                  —  IV.  •»•
                                                                                                                  o    •
                                                                                                                    •  c
                                                                                                                  c
                                                                                                                  
                                                                                                                                                                           o    •
                              •  o
                             c
—     —     •=    j  3-    if   c >
'VOX    -C   IV If

C  P-  O    0X0
o    • —    s    • —
 •  o          •   o
                             r1.  c
                             O  -C
              o
               I
                                               (Vi
                                               IT
                                           X  C —
                                             n
                                                                                                                                    x —    —  c  xt
                                                                                                   n o n
                                                                                                   c   • —
                                                                                                                      GO    (Vi
              — I*! i
              IfX—
if  e  ,1

C  P-  —

 •  c

 I

IV  f  C
(V,  x  <•"•
G   • —
 -  o
3
                                                         o  ^ a
                                                         o  r-  <»
                                                         f  — !f
                                                         <\J  O O
                                                         c   • *•
                                                           •  o
                                                                                     ru (\j f
                                                                                     J  X  (V-    r*

                                                                                      C IV •«•    X
                                                                                      — -r n.    >•
                                                                                     o
                                                                                      I
                                                                                                                      o -*     s
                                                                                                                      C- rtj     <**
                                                                                                       o  *    »

                                                                                                       c  —    -c

                                                                                                       c:          •
                                                                                                                                    I" O
                                                                                                                                    cc
                                                             — if.
                                                             e •"
                                                             C. C
                                                                                                                      — 'f.
                                                                                                                      O (*•
                                                                       C  I-  <
                                                                       —  I*!  Xi
                                                                                                                                               •  C
                                                                                                                                              o
                                                                                                                                f»  x  -i-
                                                                                                                                If  IM  C.
                                                                                                                                C  •£  >-.
                                                                                                                                C    •
                                                                                                                                 •  C
                                                                                                                                *  n e
                                                                                                                                •£  x *
                                                                                                                                '
                                                                                                                  IT.  C- «     —  
                                                                                                    c   • —     =   •
                                                                                                     •  e         •  c
                                                                                                   O — CI     C  IT  (M
                                                                                                   (Vi C ii     —  i  —
                                                                                                    * c o     <*>  v  *-
                                                                                                   tf> o —     —  <*  •«
                                                                                                                                                            —
                                                                                                                                                            U
                                                                                                                                                            c
                                                                                                                                                                          a
                                                                                                                                                                          u

-------



















-
c

1
HJ
s
^

U.
C
i
'.u
X
^
~
^
c
»
*—

a
—
2

2.
-w
~
^
"^

a
_

A

X
C
: 2.
X
t/1
^
^
UJ

id
u.
t
u

i
^
_j
.jj
^
a:
O
°



















i


e
ct

«a



^
G
a

(Vi
I/I


c
?
If.


3.

I/I



•»
C
^
-*
IT

f"
c
z
^J
J>


_
c.
2

^
^5


^
G
2
2
3.
*/>


G
*
a.




c
tr
in


c
~
.*}
~


1*1
z
^
f*»
O


_
~
I
^.
:*l
~









C — IT.
<£ G IT
^ c ir>
— c cc-
^v ^
• c
G

c CT a I
o if  c r-
* C G
• G
G
t

_ _  G If*
.» c »•
c . '
• c
-3 ;
1
G — (V,
— — f* 1
4 G f-
X C It
G • —
• G
G
1
CC G h-
C G ^- i
if. G r

• 0

^l

-C — .<
X

•*
c
^
o




2
a
u


I
-^
c
v;


;*
a
z
nj
^


<"
'G
JX

-•VI
^










G CF 1^
»•• »» G
c  c
0
1
-* G -»
•^ G IV
J) • —
• O
0

— — _
*- c *
m o tv.
(t C —
^t • —
• C
G

^ — —
— G *
"• G nj
it G —
in • —
• G
G

O «•• G
•» G C
IT. C C
G • —
• G
G


!/> — C
*• C G
ifi G G
G • *•
• G
G
1
•X J> C
fU <7 —
G r- c
G • —
• G
G
1
.0 — T
•» C —
E O I
_• •
• 0
G

C — • Z
r- G ^
J- 0 —
l*\ C X
1« •
• o
3


~
^
^
1*1
3

(X 9 G
— -C ^
(Vi (V' G
G CC G
G • —
• G
G

tv o c
(•^ ^- rf^
C h- G
— R; —
G • —
• G
G
1
f- e •*
C G (V
IT. e —
• c
G
1
-f 0 —
— c —
• G
C
1

(VI — —
— C (VI
if e- —
<*T • —
• G
G
t
(•> C •»
IT C (vi
IV! G —
^ • •-•
• G
C

IT — —
tr G -*
-* O !V1
-* 0 —
Ifi • —
• G
G

IT. — —
(*• 0 -»
C" C IV
Wi G *••
iT • —
• C
G

^^ ••» G
("CO
h- c e
c • —
• c
G


*** — G
CC G 0
•* C G
• G
G
1
r- --^ c
C; ^ O
O • -*
• C
o

 a f-
& f^ c
—  —
G • —
• G
C
1
CC C -V
t* G WA
(V. C —
« G
G
1
CT G tfl
X C —
(V. C —
• G
C
1

1*! — —
C G (U
•» G —

^ C ^
c • —
• c
c
1
< G r>

. G ""
C
*
* G r-
ir G *c
(v; G (*>
• G
C
1

•» — O
< c r>
•» C d
G • *•
• G
G
1
e o »-
1X1 G (*«
— • —
• G
e

IT — O
If C i->
(VI O ^»
,£ G (*>
— • —
• G
.G

•y — cr
sen
o o ^
G G 1**
^J • *^
• G
G

tr o (*»
— C G
fX • —
• G
e-


(•» C (•>
tf! C 3-
H3 G O
• G
G

G C 1-1
— G »
— GO
(VI • —
• C
G


•» c a
!*> •
• G
O

.r. — j>
-»&!*•
G C tf)
J"* O -i
i*"^ •
• G
a

_
*
^
^
c.
iy>
3-31

— C !T
•* G O
If. G (V
G » ••«
« O
G
1
•«r — (V
»• o *
C C fl
<•. o it
G • —
• C
G
*
ceo
C C f>
» G
G
1
(V c r-
(V C vC
c> c n
• G
c
t

IT IT »
— I »-
C I <*>
G • —
• G
O
If — »
•£, C t*l
G • »"•
• G
O

* — 
VM • «M
• G
G

»- o n
(V G O
(Vi • —
• G
G



G
^
i;
a.
""


CC C If
IT. c .*>

If. C O
G • —
• c
c
1
^- C (*>
•« e it
(V • ^"
• G
G
1
*- G ^>
tf1 O ^>
(Vi 0 (*)
• O
G
1

c — a»
•*; e ^>

M 0 »-
^1 G 1**
— • —
• 3
G

r- — cr
G G 1**
C C h-
••• G 1*)
(V • —
• C
O


— c c
n: • —
• c
o


— cm
If C 
•C C 0
• G
G

G e it
— c ,>
— C G
 c «
a o *
n c, r-
• c
G



f\l G ^>
G •
• o
C

HI 0 CU
** G •£
G •

G
1

G G (\i
•O C .C
C •
G

m G (^
G G C
fk. C ^
3 •
* o
—
1
& r* (V
^O o *\t
•* C rfl
0 *
• o
c

 t*
r— -f &*
rfl f*- ^
C ^3 *C
C "
* c
o
1
« •< tT
un « —
c c -*
o o **»
c *
• c
c
•
cc *• c

— c r-
", G
G

(VI C -O
 — as
f\J C —
3* O if>
•v£ C5 P1*
w *
• o
o

tfl T «
« X. —
— c- ,0
f*f C ^~
o »
» o
c

ftj c (V
tf c 5
o •
» o
o


> * (M
2 i i
• o
c.

«X -C (V
^W «M ^
o •
• o
*-•«

w •
* 3
o

•f — 
, tf C /"
/ 0 •
• G
G

— C til
<** — —
o — tr
(vi c r»-
G •
• G
G

n ^» ^
•» c -c
G •
• G
C
1

O C CV<
£ ?i
o •
• 0
c
1
*»> C CV'
0 C -O
fs> ^ i^
o •
• o
o
1
0 f- CV.
•c c rv
•* O X*
o •
• o
o

^, — Ai
•£ '^ ^
v^ tf C\l
— — £.
-O »
» c
i


—
^
2
-jf
1/1

X ^« C
ft. oa <,
^- o ^
(W o l^
o •
• c
o
1
en c o c
O * ^"
• c
^5
1
a* c —
— C f-
P» C 0
• o *"
c;


0 -* «
^P -O ?•
c «o c
o » •—
• c
o
»
ao — 
^ ^- ^
M« »-• ^
^ •
* o
o
t
0 * ^
i? <\. C1
f^ » p*-
<** C. )K
o •
* o
o
t
_
G
J
^
ftj
^

— c? c
c ir> *
CC G <4>
(VI C 1?
G •
• c
G
(
r- f> t-
O p- O
•G G *f*
(VI 0 C
G » *••
• c
e
1
»- (V CC
— c c
. c* "~
G
1
G O —
iT c r*
Jl C G
• o "~
G


c o tr
— cvi r-
(VI G G
G • ""
• c
G
\
(V — CC
i- o a>
(V G O
— * —
• G
G
)
•a — 05
— G X
3D O >••
'^1 G O
(V • —
• G
G
t
O — S
^ G CC
-T C ^
1*1 G C
ftj • •"•
• O
0
1
^ ••" If
ffi c 3>
c •
* G
G
1

* ^ »
5 = £
* 3
G
t
•^ •"* rf»

(f\ ^ Qh
O •
• C
o
1
tfl ~ *
>» C i*-
33 o r»
o *
• o
c:
1
pw ^ ^»
^~ ^ ^
C -s **>
J* e ^
o •
• o
o
1
r-
0
2;
•^
rtj
wrt


h- ^- ^
rfl •-• ^
tvi c a
c •
• c
c
i
O >£ 1-
•* a a
f* G U
1*1 C C
c • ••
* c
c
•
v* «« 3

IT G C
c • *•
« C
C
1
•a c -
c o r-
PU 0 C
"• 0 *"
G


o — a
tv rvi ••
c a c
o • ««
• G
G
•» — • T
-coo;
•* G G
(V • —
• G
G
1
n —  0 0
-^ • *^
• G
O
1
W ^ tf
2 (X O
o •
• o
c
1

tf o 
• S
G
1
«•• ~ tf
i- f\* ^
0 -» -O
^ C jj*
0 •
• c
0
1
IT * >»
a c K
<*v o **^
c •
• o
c

- J» ^.
.r — c
B — p*.
-X C ff-
V •
• c
o
t
*
5
2
£
'V
l^

-------
3-32
1
i


t
1














(/I
! •*-

; •"-
^
>
I .x
i *.

i
^

u.
i c
*
t x
f-u
i 3
"™"
Z
1
V

C
M
C
1 I
-
'• zi

i
>-*-*
^
~
~
—
A
t
< U
c

; ^
x
1 VI
*^
jj
*-.
u
u.
u.
! c
c.

-?
c
M
*—
4
-»
i*
~Ji
C
u























0

c

2

—
C
C!
z
c
c
c
Ik
a.

t;
u.

a
•>
<
u.
-i
^
z

^


a
~
u»
p-
_
z
o

a.
2
.AJ
•*
•^
c
^
-^
•_
~
>•«
•£
^
"*
K
z
->
«
^
o
or
£



o
CE
2
U



,M
C
3:

u



^
o
^
-£
(VI
™


e
~
•v.
~_










c
^
^*
o
•
c
1
P*

^
.
o


o
>y
£
Xi
Q.
•
O

3"
X.
c
^
•
o


_
—
a
•T
fVI
2

0 C"
^™ X4
— -c
R« f*
•
o


(Vi •£.
•£ —

c X
•
C


— (fl
O If
C X'
C X

c


(V 0
X1 3*
— X

O



— If

O X-
t: x
tr



if l/"i
0 X>
ii X
^


— r.
0 X
C in
o x
«
<=


C X
o x-



— p-
0 X
C. I
^



o -f
G- (Vi
o r
^



— i*-
c x
= rvi
•T ""T
^



— IT.
C X
c X
o re
•
<»


«•• XI
sr X
0 Xi
c X
•
c










O
ff
f*
o
•
o
i
<
(Vi
X
_
c
•
c

^
•X1
x>
q£
^«
•
C
»
^
C
0
o
•
c


p»
«H
^
•a
~
c


X
^.
c
=
0
1
r-
<~
^
P*
^
•
«

£
=
c

,*.
a.
it
*
—


T
X
On
*
c


3-
f*-
>f
^
'»
—


C
-X
^
/**
T
*
•5

O
c
c
•
— «


r>
^
_^
_,
/v.
*

c f^

0 XI
c a,
•
^


C X



•— -h-
c a
O (\J
•
c



G: cc
C" /\*
-, ni
c



— ^
C I
O "%J
C X
—



— IT

c X
O X
•
o


S X,
— x
•
c










f-
("
(V
c
9
c
1
•4
IT
pk
••
c
•
c

X
x
o
c
(VI
•
c
t
c
X;
r'
e
•
c


(VI
^
(VI
X
"^
a


(V
1**
c
••
o
•
X
(7
x
(Vl
o
•
3

e
c
0
f
„
P--
ivi
•«
o


^
l^.
(VI
•
o


•-I
^
IV
m
.
c


^
c
e
o
o
•
_•

£
f*
z
•
£,


^
—
2
^
f\J
z

(V. n
^ ,i
o P-
«
c


X 
O rf"1
C X
•
c


— o
•- 
C If
C X
.-•


n if.
P- sT.
O rf*
C X
•
c

P* Ifl

o


r* »-
c. X1
C 
•
c


C If.
O XI
c. t-
•
c










•r> tf c?
< — X
c X O
o •
• o
c
1
(Vi -— 
x. c o
o • —
• 0-
3
1
(Vi C O-
-* c o
• o
9
1
— — tvi
O O "O
e: c .->
IT C 0
». • -«
• 0
3


* c" ^f
P- 0 ">

o



••» ff
C (Vi
c ^
o o
c



— c
0 (Vl
= c-
c c
o


— £-.
e nj
o o
c o
« IM
c


ll
^


— (VI
o -c
C 0
• •«
0



= 2
o ••>
o c
—



-— (VI
o *
C 0
•c o
c



^K f"*>
c x.
O (VI
C 11
•
0


c a
C- Z
&
c










(V,
f

*-
o
•
o
t
f
Pfc
•^
^
c
•
c

p.
c
X
P-
o
•
0
*
X
c
„
o
•
c
1

•"P
•»
•y
•"
«
0


p-
x:
IT
o
.
0
•
XI
«T
c^
(M
0
•
3
'
cr
if1

o
1
=
0
0
o
«


i^
^»
m
.
o


~*
•"p
c
•n
.
c


••*
p^
*p>
M
*f
•
0

(Vj
(Vi
^f
•
o


^.
=
J;
^
o
u

— 
C (Vl
c o»
c c
o



•- c*
0 (VI
c ;!•
o o
_. ..


cv. o
j n.
— 0
o o
• «^
o


C (Vl
c y
o o



0 (Vl
o o
C 0
0~~



•. • rtj
O C"
o a
<=



«« (V!
C .0
c >
o o
o



w« V^>
o r
C (V'
= Xi
•
^


C X
c f\l
o ^
•
o










ur (T x
» < <
c cvi 
(VI C P"
C • —
• o
c
1
f — in
-f. P- tfl
Qk [fj ^»
0 (V,  •— •
• o
0
1

X — If.
C 0 Xl
c c —
•* c •«
• o
e
f

(v. *- in

— •= o
• a
o
t

-* ^* ?
17 0 (Vi
(Vl O O»
-£ e c;
• o
o
1

X P^ XI
p- x m
c — x-
— i-" i
o •
• 0
o
1
P- C If .
I'M O X1
* — I
o •
• o
o

a.
^
^
-a
2
">
<

C (T X
(•" — -vi o »
X- o o
• a
o
1


!f P- XI
x c m
(\i c x
o •
• o
o

r 5 xi
P- = r
o •
• 0
3

a
H-

*^
•£
->
4

X •£ X
*L -^ 0*
c •--« Xi
P-- O XI
P- O ~
•~ c -*
p. . M
• o
3

,c = xi
i- e »•
• a
^

r- — 
a~ o (Vi
0 C C
• o
o
1

a- o nj
P- C !?
x e o
• o
o
1

(V. — T
— C .X
•y o »
c^ o o
• c
0
1

(Vi — sr
d o xi
(vi o in
ocx
~* •
• o
o
'
cox
** — "T
a •
• ^
a
i
a.
\u
^_
x
^
~~£
3

C P- 1.
C C  
c o —
rvt .—
3
I
» — (?
— C l\l
»- 3 c
• o
•^


•C 0 (VI
(v; c CT.
P- o ••••;
• a
c


^- — ^
•f C (VI
P- =. O
•* •= 0
• <=
c


(V — X!
•COX*
(v o xi
xi a x

* —
C

*£ C l/f
-* ~ X
.-« •
• C
*

2.
*,
_
^
-•--*
^
^

•»•— (\i -a
If X *-
— 0 (Vi
o • •-
• o
o

x a: -£
0 * If

— 0 «•>
o • —
• c
c

m -« «
If. 0 XI
fn o o
C 0 •»
(V! • —
* C
o

OCX
OCX
c o -r
o • •-.
• o
*M


P-- M rt
(Vl C If
x. P- c
— o *
• c
o


a — ir
•X 0 XI
-^ c o
•^ • v^
• o
o
1
If. -* <•}
^ C X
X a o
(V 0 •<»
tv • •^
• o
o
1

o •« «•>
c « —
• c
c
t
com
C C Ifl
c e «•
CO*
c • —
• c
•M

if — ("
a- c. in
o c *
(VI • •-
• o
o


•* P- X.
tf 0 X!
i 0 —
(VI C •»
• a
o
t

(--* — tn
<*•• c xi
(V. C —
o • —
• o
0


(Vl 0 —
P-- O <*
a • •.*
• 9
O

Xi P- X.

(VI •
« ^
a
1
x c x
x, a x
"COX1
VM •
- c
a
i
a.
£T
^j
^«
£
->
4

P- — fl
X C X
f C 
— c in
c • —
• o
c


— CO
x x cvi
C — ff
— . (M IT
• O
o


i P- 0
3T ^ Al
•£ ~T 3
— Of"
O « ^«
• o
o
1
~- P- 0
0 .» IV-
p» •» »
-* c ,n
o • ^*
• c
3
i
rw <£ Ou
C • **
* C
—
I
i* —. <\.
^ C K
«— C P*-
• O
si


--» f\[ <\.
^— C P^-
« O 0
• 0
o


(\i — f\*
<*C C f*-
s-C -S f-*.
-* c- c
* c
o


•* T >C
-^ -T — •
f* tT -*
-* 0 X
*5 *
* c
•3

•C \/> ••£
f^ ^ -----
X: ^3 *
— •— —
O •
* O
—

i.
,"•
-
c
o
c
---u

-------
3-33

















t/1
2_
w
*"
<
^
ULI
vr
x
c

u.
c

3
u.
i
^
~
•^
•x

c
*
c
^
^
~
"£

a
^
ii
3

a
—


3
^
a
V

<
u.
^
•M
T
^
X


a.
^
UJ
*-
2
M
^
C
a
2
u.
t-
X
<3
T
^
3;
_
>—
M
Z.
•>
<
a.
^,
^
*1
~
->
• — rr
e c IT
x c x
(•* O CVl
m « —
• e
c

r- cc cc

C7- (VI •*.
— 0 (VI
c « —
• o
^

3 c» s
rvi X £7
IT ~ .C
C • —
• o
^
«
J1 0 X
X -G •?
V  •£
C rv (Vi
c • —
• 3
c
i
U\ •* C
«— • tV C*
o •
* o
^5
i
If . (" 0
PCS
fm ~* CT-
3 •
• 3
O
1

^ tfr &
m K. c
^3 ."• [S
0 If. >
O »
» 0
c
1
»• (VI r"
If' ?- if
rvi o r-
c •
• o
^
1
O 3 r»*
•4. if. U~>
CT X £
«— • — P-
C. •
• c
o
1
£
1
T
c
o
X
•-
— c —
c e f»
X O (Vl
• c
e

r- ~ if.
— C (Vl
— e cr

c • ^-
• 3
3
1
— — CC
c e »-
0 0 OS
e • —
• o
3
I
1- —  ~ m
u: m tf-.
e x •*
c • -•
• c
c

c — c
!•• -I C
•— C •*
C • •-.
• o
3


f -c m
(V: ^- if.
— 3; —
^» — -4"
3 • —
• e
c

— ~c in
i^ e in
rV r*> •-•
— — T
e • -*
• o
c
1
P- 0 u'l
c e '^1
e ^ ^
c • —
• o
3
•
If (" Ifl
(V. r- if.
•C in —
3 « —
• C
C
1
c>' -c ^
e -» 3
• 0
3
»
-c o rvj
7 £ £
O -£ O
C * <"-
• c
o
1

?*•?*• !M
X J"> *C
ff 1 'T'
^ -£ 0
e • —
• o
e
I
c- -c Ji
c. c vn
3 0> X
o •
• c
C
t
^r 
•— <" i^"
^-t -™ j^.
cr c* CL
o •
* c
2


3
_
^;
x
^
1/1
x c -o
*» e o
if e rv
• c
e
I
IT — T
- *- If!
<»• Ifi «-i
o ^ so
o •
• c
e
1
if* ftj ^1
•T -» ir
— . -c in
c- — -r
o •
• c
3
1




^A
^
<
»- C *
»» e c<
c* c rt*
• c
c
1
cc — ^»
r- c c
<. e i*i
»• c •*
(VI • •"
• c
c
1
c. IT n
•< r- if.
— — • -*
c • »~
• c
c;

C ^  •»
e • —
• o
c.
\

•^ «* If
~ CC If
tf! r" —
e x •«•
c • **
• c
c

If. (•> If.
(*> •» Ui
If - c
C • f
• o
c


— * C (VI
X •» *
^ .« c^
o -c o
c • **
• o
c.
1
if e rf»
fVi (C If;
o » co
0 •
• c
—
1
(V» (V 1/7
^f 't \f
—~ £ J^
C "Z ^
C •
• 3
~



J
VI
.U
t^
<
— iv *;
c» e ^
e ir (Vi
• c
c

r- — IT.
sex
c. c »
^ o ^
c • —
• c
e
1
<. « —
* e ~-

c « —
• c
e

e 
cc c ^
e • — •
• e
c;
1
4- p. —
•* O —
e> *~ ffi

e « —
• e
"Q

cr  —
*x *T •"
a xC X
C- (Vi (**
e • »••
• 3
e
1
>r. » —
^ ^ —
-* r*. cc
c 'n ***
e • •-•
• 3
a
1
0 •» —
X e —
_-. _ E
3 • —
• e
3
f
— in —•
^ 3 ^«
X fl vt
^ . —
• 3
c
1
.C C <7>
» e -c
n c o
• ^
^
I

2. c2
(" e 3
e • —
• e
c
i

s rvi —
r»* c o
^30
ri e c
3 • ^^
• 3
3
I
C« — r~.
<7 £ (^
.•V 3 3
C •
• O
3
1
If . 3» "•
C 1^ li^
3« 3 >n
(V c ~
3 •
• 3
—
1


=
1/1
^
<
'A.
C. C -f
ceo
if e (V,
• c
3

(Vi — -«
O 0 O
•» e (v
*". 3 •»
C • «M
• C
ez

cr f" ir.
-- CC If •
e ^ •«•
c • *-
• 3
3

IT f- CC
(Vi * C
o »- ^
e » ~-
• Q
3
1

» ^ Ifl
-» cr if
•^ if »-•
e> cc •*
c • ~*
• 3
0

-* — If.
e cr- in
X I*1 -~
e *^ •*
c • ^-
• 3
C

c -c .r
•c >- ar.
— • •"
c « —
• 3
3

r- *- >c
IV. O 0
• 3
^
1

]f 3cf
^* ^- 3
0 • —
• 3
0
1

•C if. (Vi
•c cr •£
cr *** o*
^- 3 <3
3 • ~^
• 3
3
1
*. — If
c-c.tr
-* 3 ^O
0 •
• 3
3
1
•f, 3 if>
^ — \4*
J* C J>
f 3 X
3 •
• 3
3
1

3
^.
>•
X
<
^
OCX
IT 3 *"
(V e (V
• e
c

c — «•
CC C «"
p^ e o
 >c ff
(Vi cc x
e *^ rr
c • —
• e
o

f f «
e rvi cr-
— 3 X
c o> ("
e . — .
» c
^
I
-»• — ir
*C, !*• O
-. -* I
C "i —
• c
o
i
rv -C <•>
j^ .** cr
rn f» cc
3 * «"•
• 3
Q

— * If.
^ i 1-
(Vl C C
» 3
*
1
e

— CO
e • ^*
• o
^
I

•c -o —

O1 <9 **"
— c o
o » *•*
* c
2
»
IT — C

O •
• 0
^
t
0—0
X J C
— -* ••
(V C I
^ 9
• 3
^
1

=
ii
>•
>».
UJ
^
OCX
(** «-• (VI
(V. 0 (Vl
3 • r~
• 3
C
1
K •* O>
I C 33
C C f"
r c o
e • —
• o
c
I
(C Ifi 0
(Vj ftj CL
^•t CC ^1
C iX **
c * —
• c
o

 X C
•*- — S
tn — CM '
c • —
* c
o
1
O (7 0
C f* X
— c: fvj
o ^ ^
• —
c

ru x c
C <£ C-
• C
™
1

< -S
3 X O
e • ^*
• 3
^
1

r" -»• 4-
X * 0
-C X f»l
3 •» 3
3 • —
• e
o
1
(Vl X K
— If C

c •
• 3
^
1
.0 ~ ---
0 t- u".

— rv, cc
3 «
• 3
O
1
c:
Z
^
x
O
T
»rt

< cc o
c -* —
• e
3

*»  f» (Vi
•« r- f-
^ e (Vi
3 • ^*
• 3
C

(V. -* If
c *~ e
(r n o
c r- ivi
3 • «-.
• 0
3
t

^ o (Vi
x e p»
tf? f*- O
^» e (v<
3 • «*
• 3
e
t
Ifl 3 - — p.
— X O
— »- (V
3 • ^^
• 3
3

M r> (V
c* •£ f*-
— (VI O
3 • —
• O
^

•4- CC (Vl
<•< X r»
X 
1

in cr x
0- * c

O • *"^
• o
^
1

^- f* ^
•i r* cs
f> **< 3
?V| O C
3 • ^^
• 3
O
»
c cr (v,
X (V. 3,
(Vi o r^
3 •
« 3
^
1
-» -VI !M
p- if r

^- ^- i»-
^ •
• c
o
i
^;-
>•
Jl
^r
••
 •»
C 1" «
e • -*
• 3
3

If. X J1
r*? l^ f.
•» CC —
3 « -.
« O
3

= X IT
rvj 3 ji
r* -* —
3" ~« —
• o
3

— X -O
3 r c?

• C
o
1

•» I f -C 0-

o • «.^
• o
3
1

•c I rvi
•C -* ^
X C C

3 « ^-B
• 3
3
1
•J) X ifi
r- cr if
— 3 X
3 •
• e
c

x T -n
— IV. lf

— (V, C.
3 •
• O
C

^
?
^
^
x
r
0

x e o
•c e rvi
• c
c
1
cr — •«
x c e
I- e (Vi
& Ci •&
e •— •
• e
e
1
fti »• if.
(Vi c ^
f 3 .»
o • —
• 3
C
1
a- -5
x. c o
J" 3 -»
C • —
• O
c
1

o ^- m
— ^ I/V
*• X •—
•"» ^* -»
3 • —
• 0
0
I
e -» f
^. •* in
-*• x •—
3 ^3 ^
C • —
• O
C

-* c in
cr -
^>
••
u

-------
i)




S



z.

{J
^
~
^
—
»-
G
¥
^
{j
C
u>
a
M,
tj
tn a
c >
— ••»
< u.
> a
-»
-i L.
X ""*
c r

fau
C |

±4 ^»
X Z
^ ^"
12 i.
z c
.x a.

o u.
H v*
C x
- =
— a
~ ^
' j- z
z z
~ ?
2 O.
_ ^
»_
A X
_ <

O >•
a. •«
a.
x =
5
tr ^

•7
te. c
.u Z
O O
u c:

—
^ _
p- c
4 2

J* ~?
I
C
0 »
j:
IV'


=
_
>


t
\







— ^- c
c c c
f— P* *•
e •» c
c • —
• C
^
'
rf^ ^ *™
— -C if
<<• ft; ^
o * *~
• c
£»
1
c e> •*
C *•• fVJ
o* ^ «t
S ^ z
• c
1
-* — —
— O <3
-* G !f*
• O
C
1

x »- •»
O ^ A»
A, G —
C • *<•
• C
G
1
•O <*• -»
O» *** A.
•C X  C —
£1 G G.
~» G' ~
O

•£ ~ -C
^j —

>» O C
c -T *"
c w

•* — -?
••^ o ^J
o c ^
•
»-•
Jt
u.



.f -If.
^* C ***
< C rt
x c i
*•« •
• c
c
1
•tf ^ iC
•i C —
ceo
^ *
• o
cr
l
sr • ~* a
*T ^C ?*•
if e x
5 ^c
c* =

IT IT r»
c a <•>
c •
• c
—


*"• 
~ -•!«
_• =
•



x
^
_j
c



X — -»
ecu
if. C <<
(V C *-
«• •
• G
G

9 ^* ^
^ c iv/
1*1 «
• o
0

a, c i-
»M )— ^
C X —
c *. *
G G

y •* <%,
iT, 
1 *.*
• C
c


(*t tf ^
X r -r
• c

o -» *»•
•— — * c
'i £ «^
C if! I'
C •
• 3
o


(Vi c m
— .-« ^
o •
-* °
7
AI t/* m
O P* -C
• w
o

C\J O O*
•£ X f^

O ;/* «C
= •
C
i
K t IT
^ £ £
C •€ -»
>

C X K
X' ^ X
— PV -r
0
1

>
5
c
^
~
IM
3-3*
i
;
^p^-in if. x c C-CP^I'C
<-*''C otr< ^"X ^i z
C* ^- >* c >* •* c^o ^i ' •»
CXA^ C?-» *\-CCVji >•
o • *- c» a * ^* M
• o • c • c i tf!
c o G > a
1
^ c < n, (\i c\. ^ c CK c

»« • •• c • c • ^^ • ^
• d • O • C *""
o c c • »-
1 i 0
^ ^ o cc — -* sc -* trw ; r
X ^ ^ G G *™ ^* P* U*l • ^
— < ir x c. o me-« c
e "^ — S°. ^ S ^ — • i
C ° G ° G* C ! I
1 U
•"CCf- (Vi — i~ ^-»«> Z
C^(Vi ftjCS \1"^ QCi ; ^
c^- c'°* c"-— c
• O • O • G- *•
e c G | ( ^
C-
a • — — • G • ••* ^£
• G • 0 • C G
C G O Z
II tf>
P»
x - = «; - * ^ - *; 9
SOS »CGi ^30—' —
-rcr- ccai .£<=•»•• >-
"~ • »^ fV « =. • i_ , ^r
"• o • a • o ^CC •£="=•' P-C** <
— — G <• e !« ~»=o ""
C • •" — • Al • *-*i
_* = ^ 3 _' = !
SJ _
(\jp-x »co xo-o tn
OOG (VC» CTC.O IT

G G G
1 1
J-X» X — T X — -"V l^u


O'iG (\(G-* (VG O)
C^-— — • tV| • — j
GOG
1
P-- — • IT *\. — tf" K^-tf5 O
G G O 1*^ O ^* 'fi C i/* | tL
\f C X i/".' G ^Y t^ G X^ 7L
O • ••• • ^— * *^
1 ~

(\i ^^ '^ <^J *•• ^T ft •** J* C
^ = -^ *C — *^ £ G ^1 ' ••
ifca. -a-c^ (XtG
G o o !«s
1 2.

3 «-
— £ —
i i c
O — t Z
— C o
0 1 X



— - ^
^ C f*^
c c c
•COG
G • •—
• G
G

35 C ^
^ C ^i
C • ^"
• c
c
1
« f, —
C (\i •»
— -« (V
G °tZ
c °

(V •" X
(\J » —
S ".-
* o
o
I

(M h* (C

^- C O
o • *—
• c
c
i
«4 -C O
ix* tr. **
(VI •- O

°
« --
r" — <\!
••\i c —
• o

4 -C  9 ^«
c t*- •$

G P"» *—
G • «—
a
«
Vf (VI —
^ ^ •*

c

G C »
712
3 ~



—
?
3
(•*
3



f -^5
a. c (*>
c^ o o
c* c c
c * •-•
• o
c
^r ^ *™
< c ^
IT C ^
G • »—
• G
C
1
^ fti •—
14^ ••* <7
1" If CVl
c "I -
o C

4 IfS X
— 4 —
G °--
• G
G
1

CP- P' 1C
<•» — in
m c c
G • •"
• C
G
I
•» !•) C
Ifi C f"
r- — a
= 0'~
1
IT f- -
— tVl I?
— ry —
• 2

("CO
0*  C ftl
o • •—
• c
c
1
— o »-
fXj f*^ ^
fl ^> CV«
S f^ —
G °

— « x
in x —
a f* —
« G
C
i

a3 e X
ri •« ?
— — c
G • »-
• G
O
t
(M (V C
-o ir f"
(VI X C
°. ="""
0
ao r» —
!o G (V
CVl G —
• C
=
;/! •«; c-
^> c\» vf;
^» -O w1
G a c
G • —
• G
G-


l\i G X
C f* G
O • —
-* °

•» n. rv
o >o —
"^ 0 *"
c
1
X (** *-»
c c •*

C if «—
o * —
c
>
4 •* r*


c
_^
r c.-  ru f»
.-! c n
"• o
1
1- J -*
£> O <*^
.-» ^- >J
= X £**
O • •-«
• G
.=
1

(1 — (-
& 1- ?
G • ~*
.- =
I

c ivt m
G • »^
0

't (VI A*
•— — K

G •* f
= • —
O

f. P-  —
c — \r.
^o"~
i
X ^ IV.
•« G P-
^1 G f"
^ a"
t
I- C -I
^^ -
^ ^
1


m
2
^
^_
cn



C ^ —

*•• (\: p"
A. c —
G • »«
• G
C

£ c £
ff G ^—
C • —
m ^
G

C — AJ
fti G f*1
f^ C ^*
c C. -
c C

A- C i
t- •£, !*•

» G-
O
1

tr — c
» .
O • f*
• G
O

i !»• •*
O G —
— — f.
C X-
1
AJ -C !\i
O> «< P»
— C 1"
• c
1»
«" •* •»
O ^» "^
iX ^- -T
C £  M *" *%<
l*"i C *
=; °
i



ry ^
••• iV
c •
* c
o


AJ r-

O *
• c
cr ^™
r -r
L^ C
""• c
o

*_ -4
rv, J*

^- ^- i
3 •
c

a- -c
l^ C. i
C \f 1
1

^ *^ (
C "
o
1


—
^1*
~
vj
^>

-------
                                                                                    3-35
       O     C7   ^« •<     (V*-***     (**  C if*    t** ~» (T.    l"\J »-«  ^    *  »•»  G    X  ~* G    ^  «-» O     it  •— o>o     tf  f* -c    •£ o ^    *- o  <.    <  c  P»    ^c  c: p*    OOP-     •£  c c     -»  c o     «t  o c     p- c  *    o, AJ •<.
       >-     a.   c -o     cc  G  iC     c  x> ^    X- c  C-    p- e.  &•    ^:  c  P-    <*  e ^    c  c P*-     •«»  o e     x  o C     •£  c c.     fi c  —    c a;  •—
       k*a^«*        c*        o   •       o*       c*       f+    •        ^*    •        i\j   •        o*«-»c*»"C*^*c**"C»^'
       I/)      •  c         *  C          »  C         • O          * C          •  C (V     P*  C  Aj     C A:  t*.    -* C  If"
       c     <*•   P* AJ     .-  c  AI     p»  c x*    «» ~  a:    ^ c  a?    x  o  rf*    x  e Xi    x  c     <—-c     ctvif.    o-ow
       •—     X   p^ x»     e  »    —• c:  P-    •«; c  *-    ^  c  x    accto    stcj**     x  c o     jvco     tt  x o*     iftr*—    i** <**  •—
       *•     c       <  c x     A.  c c     **  c c     »-  c e     c X  •*    e ^  •*
       >     c:    •        c*        c«*-c«—"C*—-e-        c    •        c   •        c«~-c«-*c»—c«—    e« —
       I—      •  C         *  O         •  C         • O          • C          •  C         •  O         •  O         •  C         *  C         •  O         * C          *C
       ^-c             cccceocccccc
       o      I                                                                                                                                                                      *

       Z.     C   O *»       OC rf^     •*  « (M     (T<\ltM     ^CCC\i     C X  U^    IT- CC  X!
       ^     —   r* tv.     u~^(Vi     ^caj    o —  *•  O tTi     Ct'  ^ O     tf  -^ O     C  (£ &     I1* ^  •—    ^ Ct  «-^
       x     —^f.     ^-  rv.  f-     Giro    o^-c    c-*c    ^-^-«r    —  fv x    *-.<=cr     cr^c     c-tro     ^-  rv c     ^- o  ^    —c-*
^»     _j     C»        O«        O«^*O»—    C«—«C«        C»        O«        C»*«O*«^C»^^O*«—O«*w
~2      »c         «o          »c         »c          *c          »o         »c         •  c         •  c         «o         »c         «o          »c
CX<=             c            c           c           c           o            o            c            o            c            o            o           o
~     l_                                  I                          I                                                       III

«*     Z     *CP-C-     c-*c     r^^c-o     -«r(C(v    rwcrv.
*•*     >•     3*   •&" f*~     c^r*     c  «£ oj    Gunfv    ^»iCfij    *^4f!|X    ^unx    •—•^cc     -C^OB     u^^ct     >O^CB     C'tVi^*    v>-^^
—     c     r^-^^s     <£cccc     •*  — t?    cci^    — ^  c^    tfcva.    r*-  •— as    tr(vs     <^—o     o-»c     r*  — c     x-* %  *^    rv^c^-    ^tx^*    ^;—i*-    tvoc^-     j**-*-*     x-»-»     ir*  f» •«     ciyc    ^aeo
       Z     Ct^     C  x cr     c«^-c     tvxo     e^ac    'i —  «
I     ^t     —  t%i G.     f*J  X^ O     -*C—    ^lO»-»    1*-^***^    ni*»C    -*  C O    —  iT. O     iCXI1"          co^-     *—tvr"     i^sc    r^co    — cc    <\i  c  x    —  (V GL    »                                                                     I             t             I             I            I             I                          l

X     C.     t?Clf>     >*-^'rf>     •rf'GPtVt    C^Aj    (V (V,  ftj    ^tf^O    <^-C    (^*—G     *C^^-     ,—  C*—     (*C>»—     t^-£f^    -*««^^
       ^     C^Oi,     Cf'-'X     (Vi^P*    ^*iT?*-    r*-^r*    •-»  ^3  C    C  X" O    (VJ<*^O     iT'tV^     iCt/llfl     *-«'*lfl     X*^*O    <<^»O*

u     >.     tvc^     <^c.f*     ftioo    ot^c    rico    (Vtocc    r\*oa>    r-oa     ^-co     ~-cc     (vco     o<3^    003?^
C     t—     —    •        cr*        c   * —    c   * •-»    c    •  -*    c*        c*        c»        o«*-*     e   «*•»     c   * -^     c   •  —    c   *  *^
~^>      *c         •  c          «  o         * c          * o          *  o         *  o         •  c         *  o         *  o         •  o        «o          » o
^:>c             o            c           c           o           c            c            o            G            c            c            <=           c
-Stllllltlll                          I

X     C     A*— **     ^-X-»     *  (M «Z.    —OX    X«—O2    ^-*i/l    tTOtf>    ^*^tf:     ^-sjrt"W     >J^M*V     ^CAj     tT!C!/»    O^ifv

3^     "^     *^   ^ '^     •"•  J^  ^     lJ^  ^ ^*    -"W CD  ^~    O C  **•    ffl  O  X1    U^  O J^    ^  O 1^1     tT   t** ^     X  O ^     t\.  ^^ ^    ^  f^  *^    «-^ ^J  »•*
jj     >-     A.GO    ^O2C    f*^CS?    ^O'X     —  C> O     *- — C     (VCO    —• C1  ^    ^-^-^
O     il     C    ^        O   *        O   • •"»    C   • ••*    G   •  ^^    O    •        O    •        O   *        O   « r+     G   * •**     G   • f+    O   *  ^^    G   *  •—
^I«t      «O         »G          «O         «O          «O          »Q         *O         »O         *O         »O         »O        •«         «G
~£O             o            c           o           3           c            -o            a            o            o            o           =            e,
               iililllifil

Z     ^     f\j   j~. j^     r-cT'ir     ^CT1?^    ^-.^;r^    ^-^»<*j    ^,  ^  f    IT.  ^f*    o^-r*     cry*-     if, —• —•     <  *« ^^    »-« u*.  —•    c-*-^
—     (/I     !\l-^-»-     i  ^  •*     ^J*«™    .Ct***""    C  ^1  •—"    P*  X  Xj    G^-ifl    -^l/iX>     f*^CC>     -»«— v>     •iGiy     «»G^*    X G  —•
       *••     ^*  (^ ***>     ^j  ^  (^     i£  ^ \f•    Ai t^ vil    (V CC  \f*    lO  ^*  ^    ^  C (^    O  •i ^     •£  C -O     ^»  G >«i     <^   C  "O    X  f*^  C^    ••• ^^  >C
A     4     *•»  ff ^-     *^  A«  P*     C  ^! O    C P*1  C    *« **  G    Al  G>  ®    l^  C ®    Aj  C X     f^  CO     t^  G O     ^  O  O    ^» G  ^    Ai O  ^
       ^     C-    *        C   •        C   • —•    O«    ^-  •£ (**•     C1^*^     ^-f\j»^     A**-«~-     XJ^*~*    ^•'O*-*
V     «4     f»-***     fV^--^     ^P-»«    -Cst^    OAi^«    v£CX-    iTfcPXf    A.'lTir     -*C:a^     r" — O     CC^    Xi—    ^-^-^
       it     ~*  ~* ^     ^  tf  t*1     fit  ^- rf»    Aj X  !/^    ^ X  rfl    ^  -^  ^)    *6  **^ f^    CC  ^  ^     ^  C* W5     «^ G «O     C1   ^  O    X  A*  ^£     *"• ^"  X
VI            **  ^ P*.     ^*A*»-     C<*!C.    CP*-C    ^.^.^    A.CX    AJ  O I     •—  C  X     **"CO     (*"-CO     -*CC    — C  (^     AjCf**
^            O«        3»        O«—G«^-G»^^G»        O   •        O«        O   • «-•G*-*G'»—3«-*O«-^
Z             *—         »c-          *o         • c          »c          «o         *G         »c          •  c         «ct         «o         «c         «c:
j^C^G           =           OOG30CG             ^G
~                                                                                  Illftlffl

*™     CI     **^  -3  y     ^  ^ •tf'     •••  f* t5    ^* •^  X    ^ AJ  Ct    C  X  U^    A*  1^ C ^* t\J     -^ CP  Al    t/"1 •*  J^     X  AJ  \f
Lt.     y?     ~CC(\(     U".  -*  A*     CVif*>X    P*--*X    *^AiX    •£  rf"  tO    XCD^    <*-»^*/1     X*'*^     -*-P*<     —"-»••£    «-^XlP     ^ifXtfJ
U,     u.     m^j^,     cp-tf*.     (t^^    r^x-f*"    Xtff-    A.-JJ"    —  XX     AfAiX     j«<^    f* —      AJA^'Xi
^     ^.     X  O i/1     ^  I**  rf%     •£  ^ ^    A* ^  ^    ^<" ^  P*    Xl  X  X'    *••  *O X    •*  jl X     X1  X ^     O P* t^    C?  ^^  C?    *-^  i/l  ^^    'C  LO  *"^
_j            ^.  ^. rw     ^>^.^.     C-*C    ^-. —  C    C^O    OlfiX    C.U.O    O**1O    ».^-^r     —  S-*
^,            C-    •        O*        G«<-*O«^*O**-«C*        G*        O   •        O*-»O*-^O««-*C*«-«—    •--
a.             •  o         *G          »  o         « o          • —         »  o         •  o         •  c          *o        • o         »o         * o         • o
^            S             C-OCOCOC-^CGOC
C                                                                                  I            f              I             I                           I
U
       c     cx-<*     o-  *£  *t     ^mx-    XAJX    P*-^^X    '•"•TX1    ^«x-    c-^X1     P-KC\J    ^otvj    x.  ?AJ    tr***Xf     p-cx<
       ^.     -^X'A*'     tfOA*     P*  A/ X    r^-/*>£    -J^X    P*-AiX    *-*^X*    AjXX*     X^><£    ^»^**^    ^^^    Aif**J7     Oat/>
       ^-     trtx     P-^-X     X-CP-    r-xp-^    XXP*    -4p*ir    ^.^-x.    cxx-     ^xo     s»i<;o    ^,^e    •«.  en  *^    p-c~
       ^     C-*P»-     GJ^P-     *—^»G    — —•  e    •— C-  c    G«X    C^X     OiyX     O  -O O     OX3O    O^*C    O-*-*     G~*-»

       oj      •  c:         »  o          •  o         » c          • c          *  o         *c         •  o         *  c:         • o         •  o         «G         • c
       i/lOGGOOGOOCGCGG
               i              i             *             i            »                                         t             »             *             »             r            I


       ^     *—   — ^     ^-CA*CT    A>£(7    JCC?    XO^    f^***^     C7^     CCAj     ^.CA*    •—  O  !\f    ^CP*     CC**

       "r     *C.>-     ^CP-     fw  O C:    ^-A.C    »TCC    —  A» X     COCT     C   *»  X     »*OC    f'GO    ^C*G    f^C(^          w    •        ^   •        G   • •"*    O   * *•*    C    •  ^^    G   *        ^   »        O*        C    • ••*     O    • ^^    C   *  ••*    G   • «•"     G   • —•
       c,      •  ^         *  c          •  o         »o          «o         «o         »c         «c          »^        »d          «c         »  *         •^
       •*OOOO           =           OC.OGOC,C            O
       i      »              I             I             I             I                                         I                                                       *             I

                                                                                                                                                                      a            a.


              CT             OCLO'T'COCOOOX-
              iZi.i-iicaOECrcr'a^

              •^•tf'fXAtAifVAiA^CCC^^
              i/>             ^*            i/l           ^>           Ul           Z            Z             ^            vj           (_}           O            4            **

-------
                                                                                       3-36
c
C
—
>-
—
»• if
.»  x —
G  if- •*
G   • —
                     c  i» *•.
                     —  -c if
                     —  cr —
                           — (v if .
                           — o  if.
                                         .«  — »«
                                         —  c J*
                                         •»•  o if;
                                         -c  G —
                                         c:***
                                          •  c
                                         o
                                          l
                            G s.
                            o c
                                                c  i~  —
                                                G    •  —
                                                       (V;— ir.
                                                       
                                                       r-  G —
                        e
                         I
                                                       if.  *- —
                                                       —  •£**•
                                                       -»  iv >c
                                                       .-..._
                                                       G   « —
                                                              f    ivisa    ?  e  *    ;?  G a>    i'    —    -c  fv.- —    -  G if •     if  — •—
                                                                                                                      -»    if-«    iro-»    r-  IT 
                                                                                                              c
                                                                                                               I
                                                                     G
                                                                      i
                                                                                   CCO-    f
                                                                                                                                          IT.  Cr>        G
                            C

                            l
                                                                                                                                          C

                                                                                                                                           I
U.
c     G
                                                                                                              (V!  — r-    -.  —
ZC
^^

Z-/>
                     G
                      I
                                                       G
                                                        I
                                                                                                                                           »O
                                                                                                                                          O

                                                                                                                                           I
                                                                                                                                                        IT  — .
                                                                                                                                                        e  c <
G

 I
                                         G

                                          I
                                                                                                 «G

                                                                                                G
                                                                                                                                                                      o  o <\*    ^;       CGJ>
                                                                                                                                                        O>GO     GCC
                                                                                                G
                                                                                                 I
                                                                                                    —  o    i  —  
                                                                                                                                                        ccir
                                                                                                                                                        GCO
                                                                                                                                                                          — r.    —  —
                                             -«x    c-  IT i
                                                       O

                                                        I
                                                                     O

                                                                      i
                                                                                                O

                                                                                                 I
                                                                                                                                                                         — ^    a:  ^. — .
                                                                                  ru — «
                                                                                  ixc-o
                                                                                                                  G-»
                                                                                                                  G^
                                                                                                                   «^^
                                                                                                                  G
                                                                                                                  —  •»    cc-»
                                                                                                                                                         «G          «G
                                                                                                                                                        GO
                                                       <.  — X
                                                       J"f"ifi
                                                       —  X —
                                                       CE-*
                                                       C*-^
                                                        »G
                                                       C
                                                                                                    c—
                                                                                                    G  -»
                                                                                                                                o-»     —  o -r     UGG
                                                                                                                                              G*    ("G*
                                                                                                                                               *^     •M**^

                                                                                                                                              G         »C
                                                                                                                                t\i —   >^m_     c— ir    G  — if,
                                                                                                                                — -»   rv  c  •«•     GGO    *co
                     x     •co—    x
                     ^     .c  c  —    r
                     T     — • «~  r    <:
                                                              G   * — *

                                                               «C
                                                              G
                                                                     —  e  
                                                                                                                  G —    3-=—
                                                                                                                                                                      GG»
                                                                                                                                                                     «^»»«

                                                                                                                                                                       «C
                                                                                                                                                                         — a»
                                                                                                                                                                         G—
                                                                                                                                           «C         «C          «G
                                                       u


                                                       0.
                                                       u

                                                       c

-------
3-3:



I/I
<
I
uu
I/I
X
C
U.
2.
Lb
C

a.
o
O
>•
^-
u
iKlHluHj NAOS'Jian Out>
*•
c
z
>-
*u
o
>
«*
c
1—
<
Uk.
k_
<1
a
f-
^c
c
c
1
-*
IX
c
c
1
Oi
o
fU
o
»
c
•
t*
^
c
o
o
c
o
* *-
X
•
•
(M
(^
a
*
O
fVl
0
•
G


•4?
r*-
c
^
•?
•
o

T
^i
»
^
•
G

K ^
^ ^i
c —
c
— ' -*
G ^
C •*
O
^ •*
G 4
• 
c c
• •-«
G
c *
C 1^
O T
• ••»
e
O P-
o —
o •*
— -»
o —
0
e. iv
C !*
O -*•
~
C P-
3 —




C P-
c —
o ^
• —
o


— !T
0 tj>
O X
C f
^ ~


(V.
c
c
o
cv»

c
IT
•
—
1

•c
f«
^
G
•
~

-h
K
~
r
•
^ •

< —
•£ «
(V C
c
c f*
C (Vi
•• c,
c
in —
c
c —
o m
C 
o
C ^""
O IV*
c
§ £
o
» -i.
O
-*• in
•» 3»
!?• A,

O IV
o



o r"
C M
O T
« f*
C


_ -*;
C iP
0 !>
0 -M
^


—
1-
O
•
e
o
•
o
o
c
c
p-
e
1
a-
IT.
•
O
1
C
(V.
c
ft.
•
c
X
•
o
•*
cc
p-
•
.*
•
0
f:
(VI
IT
•
0


_
IV
o
o
•
c
1
•r
~
*•
c" ~"



(•< .»
c •»
X •*
• —
o


_ J.
c s*
o —
= •*
(^


,x
P".
o
•
0
0
c
o
c
•
c
•
o
o
0
c
 —
— CC
-» —
• •«
c
c —
c .»
c ,»
c
o «r
o
cr —
O fj
O
e *
c
C P-
c —
c •»
o
0 »
o 
-* X (?
— -C X
c 
r\j
IT
•
o
•

5
^
(M
(V
•
O

(M
^
M*
^
*
-^
1
fw ^
c «•
r> P-
•
P- •£
C
X 0
If C
c
— IT
C X
•
O
c <*^
o •«
•

•
o
— p-
e r>
G i
•
C
— <:
c e
0
c -O
o o
o ^
•
o



o o
c o
o ^
•
a


— .0
— o
O X
•= r
^;


*
C
C
•
o
"V
•
c
— 1C
s ~&
c- 
c
•
0

f
If
z
*•
•
3
1
p- a
if —
c —
• «H
O
e *
o o
o <•»
I"
o"£
o
o —
<= -»
• *•
cc c
•5 IT
• .•*
c ^~
c —
C
O *-
o —
• «*•
o



ftrf ^
<» •—

* — •
75


-. X
C1 T
5 X
C '*''
C3


—
»
O
t
C
•
C
»
c
»
c
5
2
•
0
0
o
(V
<.
^
4
IV.
o
-C
c
o
(T
O
o
o
•
c
1

IT

e
c
•
^

rvi
_
X
^
""*
^
I
!? ^
C 0-
rv; *
c
P- -»
f c
•
o
o
•
o
*£

•a o
•
o
^ (VJ
f^ O
If I
0 -T
•
0
•y (V

c -c
w


C ^ ^«
(V G U"*
If 0 — ,
• o
o ;
i
•« — — i
If C ^ '
«* C. •»!
•* O «l
c . — -
• c
c
C • — '
• o
o
1
C f —
C — IVI
— !\i IT
• O
c
1
p-  0 *
o « — «
» o
0
P- 0 >
ceo
<** o *
0 • —
» 0
~t
•f X If.
— C d
o • —
o
1
o tvi -*•
C • "^
• o
c
r- o *
(V X -*
• o
o


~C — 
(*}
c
o
w








<7 
-
•
o








c •*
• f

X •*
— c
-J- "^
C7


                          o
                          a

-------
      i-


      2
                                                                                               3-38
             0
             c
1      I,
!      -     ^,
                    I

                   
                                 —  -M  -»
                            p.  —              I



                                          >£   an c    —
                                          —  * —
XX    ^O3?     —  CZ>
CVJCV    *«CCVi     CV^Cifl
OO    SCf"     If,  C *
                                                                                                -o
                                                                                              t IT
                                                                                                     M i
                                                                                                     r   s.

<**  c*  ^
-
                                               -C  X   •*-
                                                                                   M  c ^;
                                             -o
                                             a"

                                          — >0
                                                                                                                  ,X £    i~-  =-  c
                                                                                                                  -C if    J"  —
                                                                                                                                                               •  c
                                                                                                                                                              o
                                                                                                                                                               I
                                                                                                                                                               C  (V P-

                                                                                                                                                               iT  CT T
                                                                                                                                                               i   C  "*
                                                                                                                                                               sj  (\  -a
                                                                                                                                                                Z.  —  I-
                                                                                                                                            • c

                                                                                                                                           "J
                                                                                                                                                                                          CT
                                                                                                                                                                                           I

-------
3-39
f'HOU > !M| UMI.HK HO|PnO«0 / NUHBFK QF
\>
1 X  C
» ^ s—
1 - 3-

i *"*
t u. zr
i hkt w^
I -
! £ -*
, ?

i <— c
! ^- —
J «3 —
— ^*
,.,. jj
1 X —
1 2 21
C 

i u
u
I o
t —
j ~
! ""
' -jj
~-
t ^
; '
i


i

(
i
,
i
X — (\J

-« c <>
f\» t C
<\J " —
• c
c

T C 
O h* •£
O *
• C
c

— — c
^s if! T
tT (*» 3
f\. C *
» C
3
1


PH
~
^
"3
M — (VI
r*> o wj
(^ C i>
-«* c c
<\f * ^^
« G
O

•coo
^ c ir>
^ s -^
v* •
• 0
c
t
•£ — £
4j
• c
o
1


-*•
c
^_
-*
\
•«• <^ If 1
C 1^ tfl
***—."«
c f^ -*
c: • i-^
• o
o

P* C —
•* o e
^ O tfi
M •
• O
O

(V — 0
if C <>
<*• c r
•*• c o
o * — «
• c
^.

o -* r*
^- *-» ^
< cr ~
C tfl X>
c *
• —
3

(*> M ^
(X fV ^'
j* cxj r
C i t
* 0
s
t
a.
z

<
<
z
^ f<+ \f\
«i' C U^
^ a *-
< C -9
O • v^
• O
CT

-* c »-
•» C 0
c -= tTl
CM »
• 3
O

,/J — —
Ceo
X 0 X
•* " f'
c - —
• o
£5


—
».
^

T ** tr
-* -c ~
0 tf^ *
V * ••«
* o
o

C c: —
P" C O
(Vt O tfl
nj »
• o
o

*• J\4 O
^ C. t>
^- c i
f*^ C ^»
0 • -^
* c.
o

ec c >-
o .-* ^:
^ (X. ^
0 > 3
O •
* cs
o

r^ j- r>
^-* C pw
•& •& t
<~ iT i
• -
O

^_

i*v

5*
<
tf» (V if>
>^ ix tf;
X 0 -^
(V C ^>
c • *^
• o
o

-C C —•
^ « c
«£ e vfi
•-* *
• c
^
1
o r o
or <> o
a1 & 
if. X5 •&
O *
• C
c

t« z m

~ * V4
* o
^
1
C* ^ (\J
iT IT *>
r\j •- ^
o x x
o *
« o
c

x »r> ^
c ? .*•
C ^ X
1\* C. I
* c
c



i
—
M*
£

— f* tfi
f? C ••"
C f- ^*
o • —
• o
c

0 C ~
a; c o
c c \o

* o
c

,£ pr- 0
a: rv o
— *£ X
*™ — ^?
C • — •
* ^
^5
i
o: e N>
— ^ c
C X ^^
o » *
er •
• o
0

X. — .-)
j x r^
Jl 0 13
c <«: a:
• cr
•3

i

tj
^t

i

li"< C C
^! C (VI
J e •*
C • ^4
• c
o

(v; *v» (v»
c ^- •*
c •
• ^
^
1
•« — •£
r* c. *
-O <= 3-
•a c (^
^^ • ~*
•• c
c
1
•
-ceo
•C C (Vi
MOD
f*l •
• ^
o

» — n
<« c —

•*• c O>
• o
o
1
^

c
^
^
u
= o «
(" * •£
m o o
(V, C (VI
s • -^
• e
0

c •» •*
c » »
0 •
• o
e
t
r- r-  — X.
if1 
-------
                                             3-40
Jl
I
u.
c
X
Lu
V



•



a

o
a:
a.
V
if)
*—
•3
U*
•^
_
i^
^
OJ
*
(J

z
c
w
t—
t
— '
*f"*
2
c
u





















_
1
c
c;
a


^-
j;
~
UJ
-J
_j
c
a

~
~
^-
c
o
i
o

>
4
u
u
c
t^.
3
•—
jj
r












£
c.
0
•
~

f
(«
c
c?
c
•
—
1

Jb
>c
r+*
ftj
•
o
1

•—
X
(V
•xt
— «
*
c.

(N,
*c
•^
•
^
t



JJ
(£
<


c 
«
—









(V
•r
c
o
*
—

-c
^
c
c.
o
•
c


•c
tr.
c
(V.
*
c
1

t*
*_



t/»
aJ
*«£
«f


na •*
J •*
• •-•
o


-c •*
(V —
•C 0
cr &\
*
c



— • M*
C f»
c *-
o -r
c



•— !/}
o -*•
C f^
O *•
^C (T
.— X
p*- rr
c



r** *-
7 — i
^ *Vi
0 35
*
c


— ^*
C U.
•
—









m ^
h- e o
o c o
C 0 <*
c • ••*
* c
o
1
ru -» oj
^ *T ^
*»- \r *• r^ — *
?- C r
I" C •*
3 • ^*
• O
3
1
»-."«-»
(V, «^ -<
O ^C 3
— -* in
3 -
• O
C


3 — —
>>- 3 !»•
r o —
•» 3 -»
• 3
O


Ai — • If!
IT 0 
C •
• O
0
1

5
>
2
<
Z


PU
C
*—
o
»
o
1
tr.
tC
^
o
c
•
G


-4-
•C
•c
ir
»
o
i

IT
f^.
f
tr
o
•
c

>c
~T
^
»
^3


O
-
t—
U.
^


(V *••
M <»
• *-•
o


•* •*
0 Oa
-* O
^- -9
*
0



M (^
C ^
— ^
O ^*i
o



— n
C Jl
C ftj
O -B
*
c


5 cr
C X
•
o










^*
c
o
»
o
1
(^
f*J
^*
*"
s
•
o


n
•»
—
•*
•
o


<*
<\i
*v*
9
o
•
c
1
X
X
^
*
0

^
z
£
o
;•
w


-4- -4
X uO
f^. (*)
* -*
o


M ^
^ f^
<. ^~
n ^
•
c



— n
c —
G ^
c ***
c



^r X
C •£
O f»»
— K-
•
O


o -*
C X
•
c










n
o
*w
c
•
^
1
c
G
•
^
1
|
f*
^
m
j

^
>
S
«•
tj
^



(V (^
• ^«
o


•c «

c



v* 1*!
3 !>
3 1/1
C >-
•
3


lie
C T
•
a










\f. -C •*
•-•«•-<*
o * •-•
• o
a
1
ir ui 4-
?r c —
O IX 0
C ^ tf>
c •
* V
-}


J* »-^ •-<
f*» C P*-
;v o •-•
C C >T
* C
C
f

X X IT

C (V <-?
— • r*i it
c •
• c
c

ir. r c
C- U* C
o *
* c
^

z.

^
I
*
o


IT
<
o
•
c

z
•C
^
•w
3
•
3


^
f^
•C
(V
.
3


(••
X
IV
IV
3
•
o

!?
C:
a
•
3

C
^
V
»~
H«
U


& -»
3 4
C •«
• ^41
3


f» 4
(VI —
.•** 3
ir m
•
3



tP **
^- P^
c —
o •»
c



e ai
»- -*
i* «
o > f
** ^1 -*
t/1 vP ^
»•
*— f^ ^-
3 •
• 0
^
1

3
-•
—
U1
1


(C
tr
3
•
3
1
F>~
3
C
3
3
•
3


(**
C
f
^
•
a


<

3
*
 ^1
•
0



*- C.
c ^<-
C J-
C X
c



r~ -c
O «^i
c —
C X
*
2


c: <,
o o
•
c










U"
IT
0
•
0

&
1C
^
c
c
•
c
i

p*.
c
-*•
o
•
o
1

o
o
c
c
c:
»
MM

-
^>
~^
•
0
t
;»
•a
0
o
*
^^


0 v*
c c
•
0


3 *-
•4* r^
— a:
^ IM
«
o



*™ 3
a <"
3 f\«
c a
3



O If
3 -»
o «••>
e. z
*
^;


= J"
c: a:
•
0









-------
I   C
   a

   x
   a
   o
                                       3-41
x

v/l
2
J,
*—.
U
i—i
!A.
U.
w
^

<^
O
o
o
^
"~
—
Jj
~
_,
O-








K
^>
O
«
^


<»•
tf»
^
»
O
•
c
t

^
c
c
o
_

p*.
c
<
0
•—
•
c
I
r*i
c
^
d
1— 1
*
•D

^T
^
O
rt
-*
*—
•*i
^



X,
r\j
o
i/\
•
o



^
o
o
X
0

_
•^
—
0
«
*


..
^
c.
0
•
0






;r
^1
•*




•fl-
**?
o
-*





—
9^
~~
-*


e
r*>
fVl
•E




0
-J
^
-Z.








t*. o —
(V O O
*• » *>
• o
3
t

0 0 <*
CO—*
c o c
C ^> tf)
o •
• o
M


4- X <*-
rf* (\i r-
C1 O W
c •
* o
c
1
<»• e ^
t •* f*
* — x
0 f^ (V
c •
• ^
^
1
*^ -c c
C i *-
o > «™
c ;? r-
o *
• o
o

u.
-U
_i
«t
o
=
0
<=
•
*••


-»
>»
<\j
•f.
f*
•
o
f

r*
**•
•*•
c-
*

-*
tfi
v
tf1.
c
•
o

»N.
X
f1^
r
o
•
o
f

c
2
u
c
o
•3
o



.4
c
c
c
*
o



^»
.
.M
4"


tr.
•*•
^v-
X




•£•
«>
O
J-









-------
                                                                             3-42
fcO     OJ
»I     O
> 3 -"
•> 3» -O
• 3
j
— — -u
,— 3 S
,-J • —
• 3
1
.»•> 3 ••»
> O 3
— • «M
> 3 -*
> => a
-• o «
• 3
O
333
• 3
t
."1 3 "I
3-33
-» » — «
— 3 O 3 3 >
«n 3 O — O O
— • -A »
3 3
— — M 3- 3 0
— 3 0 .1 M —
— 3 •" -U f~ .1
3-33 — ~> "
• 3 • 3
1 1
Jl 3 "^ "W ** J^
— 3 3 ; X —
«~ • _~ 3 • —
o — o
— .3 o
0 3 >
0 3 C
• 3
o a a
a a —
• 3

— = n
.-** c — >
3 • —
y 3 o 3 •«* a
3 3 > O O »
r~ 3 9 — — • J1
•> « 3 «
• 3 • -3
fVI -1 J — -. ^
— -• — -O 3 O
TJ — X — 3 i
— -» -« .— 33
1
^30 - 3 C
3 C — C 3 i
3 • — PU •
o 3 -a
M 3; y
:? 3 Jl
» 3
C - 2
-. 3 O
-vj •

- 3 i
J". C U
— •
o
X)
o
•
r
x

•C
^
"y
2
3
"3
•
3

O
^
•
3
n
>

-

c
o

                <  -« "v;
                M  = a
                                                                  I


                                                                 J
       T
       3
       1
                O —
                30-
x  3  —    ru  :
3  S  O-    3>  -3
     O3G
»  3  ?     03  =
 S3>~     33*
                                                                                                                          333
                                                                                                                          O  3 O
 1
^

2
                 33O     -«3-O     C3~     -03  —
                                                                                               > —  —    O  3  f-
                                                                                               •*>  3  O    333
                                                                                                                          3  O O
                                                                                                                          333
                                                                                                                          3  3 =e
                                                                                                                          o   •
                    — •>     'r> — 3-     iM —  9-      a —
                    3 "V     J^ 3 "VI     3T C  Ai     ^  3  •
                                           if C.  '"VI    —  =  T
                                           •^33     3  3  >
                                           -\ O  AJ    —  3  ^
                                                                     cor*    —
                                                                                      — -c    3-
                              3 C  A*
                              — C  13
                                        f  — -C
                                        ?  3 C
                                        o-  T a
                                                                           . —    a-   •"    —
                                                                                '  C >
                                                                                  o •?
                                                                               -C  3 —
                                                                                  a. —  3-
                                                                                  X 3  >X
                                                                                  C C  "W
                                                                                  ,"Vl   « —

•x o  c
,- 3  C
(V o  iv
                              o e  C
                              033
                              o e  i>»

                                                                                  a  c
                                                                                  AJ   •
                                                                                    • O
                 C <3  O
                 C 3  'M
                 C.   •  ^
                                        
-------
                                                                                    3-43
3*  O  —     •*>  O —     a 3* —
3*  C  X)     3>O»      «** 3 XJ
O   •        3*   •         3   •

3-  ~         O             O
 I             I              I

•C  —  .n     "vi  -» ./I      ocjn
•OOO     O  O XI      O =. .3
.O  3  J1     X!  C ;n      C- 3 J1
O  o  —     *  o —     — o —
c1   • —
                                            3. o -a      » —  3      > — o
                                            O O >      .P 3 >      ^ 3  7-
                                            31 O <>      » O •>      O O 'y
                                            3035      3 O X>      T13S)

                                              • C          • 3          • 3
                                            —            3             O
                             333


                             3-  3 -O
                             "*\   •
                              •  3.

                             O
                                                                                                                      3  O  T     .O  3  tr      33-7-
                                                                                                                      — Or-     3  2>  •»•     >  3> 3>
                                                                                                                      3  3  X*     •«  3  O      £  C- »
                                                                                                                       • 3
                                                                                                                      3
                                                                                                                                      • 3
                                                                                                                                      • 3
                                                                                                                                    3
                                                                                                                                                   "vi  —  a     •» — j
                                                                                                                                                   —  O  7-     307-
                                                                                                                                                   7>  3  7-     — 3 >
                                                                                                                                                   »  3  t!     *> 3 O
                                                                                                                                                   O   •         3   •
                                                                                                                                                    •  3          • 3
                                                                    -w  3  >
                                                                    •«  3  «
                                                           r-  -c. "O
                                                           O  O •*!
                                                           C  C  —
                                                           r*l  C 3
                                   i  —     C,   «  —     o
                                                           0
                                                            t
                                                                         X,  Ul  *>

                                                                         3-  -.  —


                                                                         O    •
                                                                                        >*  "U
                                                                                        >  3
                                            •C
                                            3
                                            X!
                                            •V)

  C      O C  £      i  — J1
 —      ~. x  —      -:  s- -a
                                                                                                                                                   33 —
                                                                                                                                                   3   « -.
i.  — .n
*\i  3  O
r-  — _i
300
                                                                                        C
                                                                                         1
»;

5
r~—jt     oo^>    fu — »*r*
C^Si     CCS     C  3 -C
LP3J1     O  3- .TV    X  C. L/1
r"3—     cc—    a:  c  —
X>   • —     C-    • —     X   • —
       —      3  C  Tl
                             r- —  x>
                             C C:  3
— _ ;>      /v,  — >


•n 3  o      r-  3 -0
-£  3 -C.      C  C .C
IX]   .         1}   •
                                            •*>  c  a
                                            •C  0  iT
                                            *  =  -•
                                                a  -o
                                                o  «
                                            -1C—     -O C -
                                            ;>  c s.     •* c a
                                             o   •        c    •
                                              • c          • =
                                                3  —     O*  3
                                                O  !O     —1  O
                                                                                         1

                                                                                        0  3-
                                                            I             I

                                                           —  3 ru     \n »- rvi

                                                           =0  X-
                                                           3r   •
                                                                          'X*  <•*! 


                                                           ^  ?•  3     J1  C 3
                                                           — —  111     »•<  O I/-I
                                                                                                       C    • —     3
                                                                                        >  £
                                                                                        ^   £


                                                                                        ^j C>


                                                                                          •  3


                                                                                         I


                                                                                        •O —
                                                                                        sr  c-
                                                                                        »  3
                                                                         3
                                                                          I



                                                              =  —     3-  3 i
                                                                             3. IT.
                                                                                                                                                   M  3  —     33 —
                                                                                                                                                   3    •—     rw   « —
                                                                                                                                                     * 3          •  O
                                                                                                                                                                  COX
                                                                                                                                                                  K  3 J1
                                                                                                                                                                  ij  — -fl
                                                                                                                                                                  •i  C  i,
                                                                                                                                                                  "V.  — J^
                                                                                                                                                                                 0
                                                                                                                                                                                  t
                                                                                                                                                                                 3-  3 IT.

                                                                                                                                                                                 •i.  ff~ X

                                                                                                                                                                                 *  J J-.
                                                                                                               £     C — C      V. —  ,
                                                                                                               —     —  O —     •>•=>•




                                                                                                                        • 3          •  3
                                                           rv, —  i—     3  J  3

                                                           ^J O  3-     'VI  31- <>
                                                                                                                                                                  Z  C —     33  —
                                      -O *\J  3     1^  —  3
                                      3 fX  J     i  C   ~
                                      S —  3>     3-  3  >
                                      rv, c  ~-     r.  3   »»
                              3 — 3
                              -v. -  r-
                              3 3 J-
^
3
-^
z.
i*5


O
"C
^
v« «• O"
•*\ c *
ff- — ^
i> C O
• 3
C-

P»- O TO
KV — ^
•^CC
>J •
^ — O1
l/> O *
— o i
*n o -c
* o
c

o c, ^~
»* r- ^c
c c- -c
A, •
r- — 3-
— c, *
^* O -O
— CO
* d
Ci

Kf C. •£.
*•» c ^:
a* »
Kt -• rv
^; O -C
<"U c ^
=7 C, J*
"i ^
o

5*5
— . — j-
O •
y~ ~* j\;
^ c -*
c 3 ^~
IT c tr.
• 2
"z

C > *
-^ — x.
c •
,0 «. nj
'V* C1 -C
c c ^
in c ir
* o
•c

**• X >C
^\ ^ u™
c •
> Ai >-
C? 31 ^*
C *O ^
3 C" i
* o
c-

0=0
^ — c
o «
ff ^v »*.
K- >• r-
•*• r- >•?
o r- -c
• -
^
l
^ = ^
r^ c -i
c *
a: *v ^
r«. « r—
as c: o-
•*• O -O
* «
C1

•* o ?
C O %ir
c •
^ *- ^?
-C ? O-
o- ^ cr
— - **~
• o
o

o- x c
AJ f^ ^
c f*> r-
•c «
E> — 3
C" — ?
e c c
^ c r^
. -
c

C. %i &
r\, c r»
C" *
t/> — • ^
•j* - J-
rf> c •?•
i c -*
• 'O
o

i c c*1
h- c ^»
«• •
-C — sr
= c- cr
x ~ ?•
c -r —
"V •
c
>
c ? -3-
J* C *"-
Al •
               -C  "V.  -C      -C —

               c  e  tf      rv, c
               iv  c  —      3   a-  *
                              fx  —  r-
                              *Vi  —  IT
                              c-  o-  i
                              -C  IX  f
                              rv*  r-  ;r
                                             
C
^

~
V
~
*
c;
1
^
c
AJ
KI
Aj
•
™ ™
—
C
— *


^ .
—
C
O
•
c
A.
•O
~



Aj
-i
"*N
C
*M

AJ
FW
r-
^
*
0

A;
CT
<>
^<
r^
•
"*
c
-
c


..
s
•^
c
•
c
A.
^
C



AJ
•—
•C
c


f^ —  3 "C
• o
Cr

•n — -v
— C *"
=» C i
« c et
a; •
• -c;
A.
tr
-
•
—

^
^
v.
5J-
Aj
•
O
—
e
c


_
c
c
e
•
c
A,
nC
t



A.
r-
c
 r -C
— c: c
A. C. A.
• C;
C'
1
5- — ^
— 
-------
                                              3-44
2.
>^.'

2
o

2
-a
z
ru
r
IV,
«
o
at
-S!
Kl
3
2
t/a
3
o?
T
or
O-
cr;
^
=
2
c:
I
a
=
3 » —
— o-o
o •
• 5
.--. 3 >
a O -7
• V
a — a
!t ^ T
3 • *•
• 3
3
.M 3 >
C J 3
• 3
0 C 3
^ 3 31
3
£. e f-
"O 3 *"»•
• 3
3
— C 3
IV 3 ~
« 3
1
C — 3T
I 3 —
it 3 3
• 3
3
1
T
3 - -C
f\. Ct 3
3 3 —
• 3
1
"V. — =T
\J* C -C

» 3 a
•xj •
• 3
3
1
i o a
•c *- _*•
c 3 .r
3 • —
• 3
1
•- 3 a
3 c 
>. 3 —
• 3
3
1
"" - ^
• 3
3
f. — c
333-
IT 3 ~
• 3
3
— 3 >
* T J-
P~ 3 X
• 3
1
IT •= 3-
I.' 3 i
C.
t
* = *> c -v.
• 3
1
3 — •*
a 3 i
-r c iv
• c
i
01 ~* ^i
•XI 3 -C
• 3
0
1
3 ,A a
— — s
o •
• 3
3
1
f -y •»
* 3
3
1
.0 i. y

0 » —
• o
1
o -* •*>
O •*> 0
0 • —
• 3
1
—. o j-.
?• s -\*
t
J 0> —
»/* .n 1/1
IT S. rtj
O f
e
i
c *
7
-c o ir
~V* C; :T
I
c — *
fX. O if*
. o
1
Ct C- •*"
o • •*
• V
o
1
•v, ^ *c
> ^ X
- «*\ o
• o
1
c c i
«** c c
1
•^ -o r
«. IT «
*. c C
• o
o
— rt '-•
o •
• c
cr c ^
• 0
I
ry •« a

o • •*
• ^
^ 3 J-l
O 3 -^
O • —
1
J- C 3
3-3-0
• 3
1
COX
3 o r»
c •
• 3
1
-C. - r~
c •
• o
T
.r c -i
o c -c
1
^ c —
j- o -e
• S:
T
C 3 X
r- 3 -
• 3
>£ IT iT
tT o ^
^ C —
• 3
\r => *•
-C 3 —
• O
0
5 iC ?
3 a a
• 3
3
1
^1 3 O
• 3
O

O « —
• O
a a a
3
3 £ «
• O
1
(v a ***
ru 3 c
./> 3 C
3 -» «
0 •
3
1
-C 3- O
3 j- a
i
0 IT -C
3 r- 3
o — 3
3 • — .
• -^
l\ 9 -C
•0 0- 3-
• C
c
C CT 3
3 — C"
— rv 3
o • *•
• 3
S ^ 3

- 3
1
•J- f ^
3 fX •£
0- W, 3
— C 
o ^ a
» 3
3
1
— • 3- a
j: 3 .n
• 3

O • —
• 3
3
« HO ^

^ 3- 3
rw 3 I
• 3
3
«*% o ^*
O 3 O
0 •
• 3
3- C* 3
•• rv
• c
c
IT 3 "•
C -T C
— IV 
1
-a 3- /x


• 3
3
£?£
• 3
1C S*
J/l "\J ™
(\l 0 *
o *
IT U- -
P*\ K\ O
^. C" ^
C *
• o
c: -s -c
— iT =3
O » —
* V
t
•= -C 3
— c T-
• O
c
i
C -C =J
— c. ^
c * —
• —
•> ?• «
- C -X
rx 3 "
X C C
c
— CO
• c
3 «O >
3 r- c
o •
• 3
3
•*> .n a
•w a :/>
3
1

1
C ~5
• 3
3
1
£ 0 *
• 3
1
^ fVI f^
3X3
3 e o
•» ru a
c. •
• 3
 3 3
— 3 s:
i
0 — -C

• 3
1
C f 
J1 -1 3
3 J^ "^
3
t
a 3- 3
c • —
• 5,
k> r\. a-
• —
o
i
Hi
• 3
3
IP 3 —
» 3
3 C J*
3 *
•^ O AJ
C 3 X
•j- — x
U* 3 *X
-t 3 i
X 3 3
• O
c- = x.
333


P^ C r*
— 3 3
3 C —
• 3
it C- 3-
X 3 —
• 3
3
a o •«
J1 3 -^
• 3
3
I
-C 3 3-
"* O 3
• 3
t
M 3 J>
C •
• C
3
-C O »
303
• 3
1
-\. C —
IT 3 f
y- ^- -^
» -" ST
— (V, •!
O •
• 3
i 3 J1
j- e ••
3 •
• G
c o- j-
• 3
•x » x
e
I
•=• - j-
3 •
• 3

A. 3 _T
• 3
ni a- n.
f- j- 3
 => a
• 3
3- 3 a
jr. 3 .1
0°
r^. 0 J^
O • «->!
* «
O
t
^ O JJ
-^ o — (
-T °
^ o r
•sr o  ^ ^ i
c c =c f
c • ;
• 3
3
1 !
^C C 0
^ ,-^. c. ;
*T C S '
O *
G
1
^r — -i,
•r O 3
• 3
1
,- — ^
• 0
C
1
rv: c ^- ;
c -r- ?•
c- • — .
• ~
i
'N. -Z ^C
•£ C -\'
« tr
t
a c ^ ;
IT C AJ
ry sc £
c ^- c
*- ftj (V ,
• 0
C |

1 s
! •*
t — *
i ^
>i

|5
! =

'=L
,' X
i •«
1 ?
|

U-
I
C
-*
t ^
1 c
1 "^
1 ^
r
C =5
•£ O
O «
• O
3
^ Jl
— ^
• 3
0
1
S ~*

* ^
t
3 i
• w
1
^ c
c o
nj o
1
O O
^ •
• o
•c o
e

• 0
•£ C
C •
• C
1
3. •*"

• C
IV —
e c
' • C
-c c
rv o
• ^
c
                                                                                             L-

                                                                                             C

                                                                                             a

-------
                                                                         3-45
       3      -* «•
.C  3 XI     3  3 XI
XI  3 •—     a  S3 ~.
3  3 ~«     r-  3 -*
                                33     !>  3  O

                                 • -«     3   • —
                            O  -•


                            •XI  3  3     3


                            3   • -*     O
                                                                                    3 — J1
                                                                                    > o  sr
                                                                                                  3
                                                                                                  ^
                                                                                                  3
                                                                                                   3 —  XI     .» —
                                                                                                   » 3  3     30
                                                                                                                O   •
                                                                                                                 •  3
                                                                                                                O
       j;

       Z
       D
^  O £>     O  *^ O     "fi  G *^

 •  o         *  o         * o
coo
 111
                                          *  C. 3-
                                          f»  O »
                                          •M  O 5T
                                          O  — Xi
                                          •u  o ft
                                          \i  o a
                                                        ^1 -.  3)
                                                        •C 3  ^1
                                          ^  O ,fi

                                          >  = M


              3   • -*     O   • —•     O   •
                .3         .3          .3
                                                                                                         .0
                                                                                                         "U
                                                     3  3 —•


                                          3          "3


                                                     t


                                          r- —     •<•»  — >


                                          3 ~U     2*  •*> -O
                                                                                                                       3  3 i\     3  a —

                                                                                                                       3  "• -<     3   •
                                                                                                                                                                         3- •*!  —


                                                                                                                                                                         35 XI  r^
                                                                                                  3
                                                                                                  I
                                                                                                                C   •
                                                                                                                  • 3
                                                                                                                O
                                                        Xt  3 O-
                                                        r»  o -C
                                                        A!  O 3
                                                        (V   • •*•
                                                                       Kl O
                                                                       -» 3
                                                                                                                  i  »«  3     ^*  ••• *A
                                                                                                                  i  c  >     r-  c  
              •V
              -w
                  c.  *;
                  o  o
              o-  c
              •~  -
              •v  o
f\J  3 -O     X"  3 -«•.     AJ  3 •**


—•O3     ^33     —  3 31
       III



•C     — 30-     -C3P1>     3  C  »


IM     IT. S3     3-O3     LTIO^
                                                                          3 !>
                                                                          O r-
                                                                                        = >
                                                                                        O ^->
                                                                                     — 31—     —3


                                                                                      • C          • 3
                                                                                     O            3




                                                                                     TOT     O C
                                                                                     r» o  o     m c
                                                                            . —     -v   •
= o-     <\t c e?
3 O     3 C {•
                                                                                                                _, — X.

                                                                                                                ,M O 3-
                                                                                                                O O VT
                                                                                                                        X>
                                                                                                                        *
                                                                                                                        XI
                                                                                                                               t

                                                                                                                               a   >
                                                                                                                                            .n  f>« -n
                                                                                                                                            3  J1 3
                                                                                                  r^ a  a;
                                                                                                   0  O  J1
                                                                                                                                    ._     =   « _
                                                                                                                                            JO  — S
                                                                                                                                            f*  C J-

                                                                                                                                                                                       3
                                                                                                                                                                                       I
                                                                                                      3 'X>
                                                                                                           -•  3 ^*-
                                                                                                                       —  -^ >     3  XI XI
                                                                                                                       XI  "^ -t     £  3 3



                                                                                                                         •  3          •  3
                                                                                J.  _.  XI     3  — ,0



                                                                                tT  3  —     3  3  i.



                                                                                C             3
                                                                                 I             I
                                                                                H>. O  T      =  t  —
                                                                                3^3     X  i-  0^
                                                                                3 3  —     —  3  fi
                                                                                —   • -*     3    •

                                                                                  • 3          •  3
                                                                                =1           C
                                                                                              X- c  o-
                                                                                              •- =  y
                                                                                              "O C  CC
                                                                                                                                                                         xi  •- a
                                                                                                                                                                         >  =  ^
                                                                                                                                                                         <^  3  j;
                                                                                                                                                                             O ~
                                                                                                                                                                             -  c,
                                                                                                                                                                          3  ~  f
                                                                                                                                                                          •Ci  O  «
       a
       r
n      •—     —
    3 o

    3 fW
              
c i
                                                                                    X
                                                                                    o-
                                                                                                             — IP
                                                                                                             e i
                                                                                                             c x
                                                    r^ —  x>    r* —  -c
                                                    331    3 3  >
                                                    -vex1    —CO-
                                                    TS-"    xi o       3 t\,  y

                                                                                        •C  •*     r\, c  >-
                                                                                          •        o   •
                                                                                       .3          mO
                                                                                                   C
                                                                                                                               S — X.
                                                                                                                               O -3 «
                                                                                                                               « C  X1
                                                                                                                               f» O —
                                                                                                                               <\   • —

                                                                                                                               o
                                                                                                                                I
                                                                                                                iC  —  X1    IX  —•  -C
                                                                                                                
                                                                                                                                             C  3  tO     C- 3  3-
                  — x     —  -c xi     «;
                  o IT     '\j  ^c ir     c
                  3 *»     c  c  •—     »-
                  33     IV  3 C     O
                                                        3  C —     c  3

                                                          • 3          • O
                                                        c            c-
                                                                                                                               X C  —
                                                                                                                               X1 c  ,-v.
                                                                                                                                                                         l\.
                                                                                                                                                                         3
                  f^-y     -cox1     -cor*.      cctv    xo

                  =  i     /vcr     ^3—     «•<=•—    j-c
                                                                       cor*
                                                                       =r  c  3
                                                                                                          t.    r~  c  {.
                                                                                                                c    •
                                                                                    3
                                                                                     I
                  n. j-     —  — ir     »>_»     c  — x-     x  —
       n.
       2
              i  — A.     r-  —   ^ A.
                                              —  *.      r^ —  -*„     u >
                            ^  3 ^     t/^ O  ST      —• 3  ^     I""* ^  ^     i/*
                            uo<>      ccr      irec-     f-?3      c
                            A:o«     fV3i:      fV'Cct     — —  I/"     X"
                                                                                         3.C     •*  
                                                                                         C3     r*O3     ^-c.—
                                                                                         3tr     —  — IT     ".  3  «-
                                                                                                                              e
                                                                                                                               i
                                                                                                   e  IT  •»•
                                                                                                   3C  3  ''I
                                                                                                   a:  o  —
                                                                                                   "^ 3  c
                                                                                 ^- C  *">    X  3  3-
                                                                                 f C  —    =  C  »
                                                                                 3 3  3C
                                                                                 C   •
                                                                                                                                                           3 C  C
                                                                                                                                                           o-   •
                                                                                                                                            (V
                                                                                                                                            
-------
       s
       .U
               3 » -•
               -u -• r-
               o o i—
               o .n »
               1.  — O
               .M 3 9
               O   •
               •*> —  o
               f» 3 3-
               — 3 f
               IX — S
•n  3  3
•a  3  —
3   •  —
 •  3
O
              3- _ _.
              3 3 .n
              •-» 3 3"
              33 —
              3   • —
               • 3
a  in 3      3  j>  o
••»  » J      3  .A  -S
^  3 J"l      2O^
333      .3  J*»  O
3  C J-      X  C  >
3  O  C      -*t 3  -O
r-  c —      3  c  —
o   • —     c   • —
f  3 /I
•"U  3 3
,0  3 ->
O   « -*
 •  3
3
                             o  x o
                             3  *V}  0
                             x>  x .n

                             o   • -»
                              • 3
                             3
                              I

                             _ _ ,-\J
                             i/l  C. >
                                           •vj  —  .n
                                           -»  3  3
                                           .O  3  O
                                           ••i  3  in
                                           3   •—*
                                             •  3
                                           3
                                            t

                                           3  P-  3
                                           i—  "1  \J
                                           3  C  O
                                           V!  3  •*»
                                           3   •  —
                                           •C  Pt .
                                           3  — -
                                           —  —  Jl
                                           >  3  3
                                            O  3  3
                                           =T  O  Jl

                                            •  3
                                           3
                                            I

                                           :>  3-  3

                                           •U  C.  X     M  O  35
                                           AJ  3  rf%     3 »-  -»
                                                                         -C.no
                                                                         /I ./I  3
                                                                         — on
                                                                         3   •-«
                                                                          • 3
                                                                         3
                                                                         I

                                                                         •S J)  3
                                                          3  -• .T1
                                                          »-  3 3
                                                          m  3 -n
                                                                        X>  3 ••»
                                                                        in  3 1/1
                                                                         3    • —
                                                                                       n  -•  3
                                                                                       o  in  .n
                                                                                       •~-  *  o
                                                                                                                                                                — in -n
                                                                                                                                                                =  a:  xi
                                                                                                                                                                 3   • —
               iT  C- 7
               *•  c  O
               O —
               3 C
^,     —    in 3  <
*\  O  3     Jl O  3
^  C  *C     •*! O  -C
3  3  —    33  —
                             •a  o  a
                             —  3 _
O-  =   -C     A.' C  C
                                             »  3
                                            O
                            ru O  O
                            •O O  —
                            3   • —
                             •C •» •«
                             J1 3 »
                             » ^ -0
                             O 3 -•
                                                          •v; 3 c-
                                                          -C C  •£.
                                                          y  o  -•
                                                   X     X  3 f\t
                                                   ^     -O  3 TJ
                                                   i/>     rft  3 Jl
                                                 , —.     iv   . —
                                                           •  3
                                                          O
                                                           I


                                                   3     \n.  = 3
                                                   •»     -O  3 "1

                                                 i —     ru   • —
                                                                         3
                                                                         I
                                            — 3  3

                                            "U 3  tfl
                                                                                       —  —  in
                                                                                       I—  3  =T
                                                                        — O  -^
                                                                        ru o  r\i
                                                                        cox
                                                                        IT 3  3
                                                                        .£ O  "*
                                                                        >" =  in
                                                                        x   •  —
                                                                          • 3
                                                                        3  C -y
                                                                        O  C K>
                                                                        3  c in
r\i  3  •*      O  G  **     -C  — --,     JI^-iJ     — 31  X


ftj  C  .n      O  3  /*.     33J1     IX  3  •**     3^13
C*-—      3   • —     Ai    * —•     3   •  —     3» —
  •  3           .3           .3          .3          •  ^
3             —            C-             3             3
                                                           I


C0^3      ,/•  — -V.     O33     ~.£ft.     -VX3
OO"*^      3 — *J     *"-3"*!     ^>3X     -** — 3
co^r      i\»oi/\     3^oin     —  3  ^     cj~in
                                                                                       ftj  ~~  ji     f\> ••• **i     in •••  tn     ^  in  3     *\j  P^ ^n
                                                                                       »n  33     •• c <"\.     x 3  3     ^  —••  ix     £  —** 3
                                                                                       -C3»**     rv^r'XJ     ^>c*^     "^£X     i.*"i/*c

                                                                                       «•*•     -••—     »««     c»-«     3« —
                                                                                                                                                                o  \r j-
                                                                                                                                                                r.  -^  -r
                             IT  C  —     —  o —
                                                                         —  C"  :T
                                                                         *  C  -1
                                                                         i  3  J-
                                                          c:  3  .^
                                                          O  3  3

                                                          e-  T  :*•
                                                                                                      —  — IP
                                                                                                      P^  C 3
                                                                                                      C  3  • c- ^
*\ C tS-
3 •
• ™
cr

J" — •**
•C C ST
"*" Z ^
• ^
c

= — f
'"V ~ ^T
C C ^~
*- e r
• ^
c

C1 C X
c o ^
C C 01
CO*
— •
ec — -v
C" C" .£
_1 o —
c • — *
="
I
3- - X
< = =r
.J- .-. ^
^ • ••—
* o
—

O — J*
••^ c ~
a: c ^


r C: IT
C C -^T
C C *-
- c ""
~. c* *"
•«

c — »*•.
A. C. ^
o o »*.
fH. O X
K* •
•*"* C 3"
5T C C
— C «•
— • —
~* ~
r
* — X
-* O ST
•*•• c; —
at • —
* ~
c

CZ O IT
C C 5
c c- *-
= _• —

> *- LT
M«i C Q
1C**
'V - —
*• c* ""
i

^r — ^>
^ e, ^
X- C f*-
** c  O £
— cr —
o _» —
J" "*
i
C C X
C O CT
O O —
o • ^
• o
_

(C -» J-
f* C- 5
^ 0 ^
f * <••
• —
c-

y- — x
* C IT
^- c r*
x c —
*• c* —
c

fr ^ f^
f*- C ^
•£ C P**
A^ C: 
ix e *
o •
— O 3
sr c •*•
•£ d X
^ —* ""
c* ~

rv •*• rv
^ — 5-
rv. =. —
— • *•
* o
c
1
J-> — a
f*^ — ?
= c *
• ^1
C-
1
X — IX
•^ c c>
— C -C
•x = —
• c
c
1
c- c <
— c a
^ "». ^
•— c ^
0 •
-c — .n
^^ c -^
x» c x
r^ ^» —
-* "

'V — rv
"" 'Z e
3- C —
C; • —
• c
O
1
7" — f\.
-C C C"
re e -r
~ J "~
i
•^ — rv.
^ c o

C 0 «
*• • *-t
* S
*

•^ «- rv
Lf C ff-
^ C X
• C

— — f\.
**^ c ^
X C C
*** C1 —
• c
—

c — ^
•5 C 3-
^ c- ^
u" c £. C ^^
C X X
* •
? e c
— ^ -~
— . — j^
~ __• —
r* ~
i
-
= _" ""

— r- x
X C* 3
a1 •*• p*
Z "^ —
*"• c* ""
c

^ C X
X fV* ^
— S, ^
C ac X
c: •
                   c o-
                   3 JC
:>  c  3
3  c  r~
OCX
J-  3  3      — 3  =T
»^  e  P-      ft. o  p-
•^  c  
-------
                                                                                                 3-47
       a     — ~-i  a     j\ •*>  -a     io
      -.     3 -• ->     -1 O  ->     .3 ->
                                                        33  3" A     3*  3  7     O  Jl  —I     —* 3  -3 •
                                                         O  3> -1     J-3-VJ     —.  3  «»     ./» O  -•> '
                                                        Al  3 /»     •**  3  .Tl     •*  3  /*     CT O  /> >
                                                         •  3-
                                                        O
                                                         I
                                                                        • 3
                                                                      3
                                                                       I
                                                                                      • 3
      2     .\! -• X.    .« —  -f>     —  —
      ^      Q .3  O    *•» <3  3     Xi  3
      ^     -VJ3—    4> C  -•     3-3
                                                 X>     f,  ~~ J\    <\J -•  -O     33X
                                                  O      a  3  .3    333     3 — .A4
                                                 —     •*•:>-•    -C3—     £ .0  >
                                                                                                   -U O  —I
                                                                                                   M O  I I
                                                                                                   O   • "~ '
              >lt>/k     3  33  X*     <>*  X»


              err^Xi     OCX1     —«  O
              •3   • •*•     3   •  —    C    •
               • o           •  C.          •  C
                                                                      3- 3-  —     -Vt -» a
                                                                      -c &  ro     c c  -o
                                                                                                   c ji  x
                                                                                                   o   •—.
                                                 <\1
                                                 .n
                                                        £   a —
                                                                       — 3  XI     -i •
                                                                       3   • —•     3
                                                                                                   •*» O  <\J I
                                                                                                   i\J =  J1 |
                                                                                                   C   • •*
                                                                                                   O
                                                                                                   •
       ^     t


       i     -v: .A.  3
n
^
•*\
J^
«•



•** "M _f>
r* r\j -j
3 — -«^
f\* O J1
— • --
• C
—
f
•%*"*• J^
yl — ^
-J J^ VI
O P*- Jl
3 • —
• O
O
1
o T rt
-C 'V ^
CX j-» -rt
G J"l J">
~ • ••
* -^
2

f*» •*. *—
-C J" •"
0 ^» M
^* ^ ;/>
^ • -^
• ^
—

J^ •" *
3" 0 ^
^3-3"
3O O i/»
— •
• 0
^
1
J\ — Jl!
— C- T i
1-4 ^ "A
*O 3 tO i
0 » — 1
• 3 i
O

                            <>  O X*
                            -S  ^ -T

                            "\j  3 .n

                              • 3
              -> 3-  XI     3  ~
              X C  3     X'  •-
—     S     A.03     A.X3     .>C

A     =<     "^TX     O—^X1     — C.
                                                          • -3

                                                         I

                                                         r  —  X
O — —    O  — U
—  3-1     ^3-0
XT  T ru     ->  3 3
•^  3 X*     3  O X*
                                                                       .•Vi C  i-Vl     303
                                                                       « C  X     —CX

                                                                        • o          •  r
                                                                                                   aj -u 3-     ^

                                                                                                   o   • — t    i.

                                                                                                    I

                                                                                                   O C  3     i
A  *O —^     *^  *^ "*     w *• ^^     O *** ~O     **  ^ •&
*"     ^»O3-

3«—     3   • —•     3"-^     3«—     O» —
  • s3          "3          *-3          »3         «3
3            o             i            e            o
                             I             I             I
                                                                                                                            *v —* ^j     *u  r* A*
                                                                                                                            -y •» 3     -~  x 3

                                                                                                                            •* ?*> x>     O  3 X>
                                                                                                                            3   •        3    •
                                                                       -. _ —     3- —  35
                                                         o   • —•
                                                          » c
                                                                                                   •u  =  J^     s
                                                                                                   e   . —.    s
                                                                                                                               •—3     3*-3     O-«3     IT-*
                                                     =  33-     — 3?      3Z>      33
                                                     •O3—     33—      r-0—     CC
                                                                                                                                                                                        —  s:     ji  —
                                                                                                                                                                                        3  "U     -^  3
                                                                                                                                                                                        ^x     oc
                                                                                                                                                                                        3  T     3  3
              3-—3-     ~-T3     Ai^-3
                                                                       -CX.-"     r»  — 3-
                                                                       •"XX     f  C X

                                                                       f\J C  —     3  O T
—     •*•     ~XX     ^~X     P-XX
U~     3     3t"U3     X^"3      3f**3
                                                                                                   »^  C  3 '    ^
                                                                                                   P^  3  ^- I    Ci
                                                                                                   O-  a  — I    5.

                                                                                                     •  -=.     i   'a
                                                                                                                            J-
                                                                                                                            C
                                                                                                                            3   • —«     C   • —

—     —     =r— x     ^-v—     3
                                              T  f
                                              —  3
                                                         > P*-  3
                                                         I "  rv
                                                                       — »*1 X     O  C  Xr
                                                                       3 x -o     c  c  >r
                                                                       3  O —

                                                                                                   r  — x i    3
                                                                                                   31  O 3 !   . tr
                                                                                                   •u  c:  —    • j
                                                                                                                                                                       I

                                                                                                                                                                      3-
                                                                                                              <\l X< C
                                                                                                              ~  -C 3

                                                                           c -c
                                                                           X 5T
                                                                           "•* J1
                                                                           <3 3
       3          33O      C
       jj     rvj  'M y     Ai ™»  Q*     -3
       ^z     cxx     ctrx     c
                                              3  -C     — — C
                                                         3 O X
                                                         O   •
                                                                       —  O »     I-  C
                                                                       CC  O 1C     ^»  O
                                                                                                                            3  -C  C     as  3 3
                                                                                                                                                        -  X
                                              —  £.     
-------
                                                                                                   3-43

              J  S  £
              O   •
                                                       >  = -s.
                                                        O  G Jl
                                                                                   ry —  L
                                                                                   J O
                                                                                                                »  G -T
                                                                                                                  O X
                                                                                                                                                        •U — J-v    _  —,
                                                                                                                                                                          3 —     ~J  C  ^.1
                               •s: 'M    T  2: ~j     *\i  o.
                                                           i C-
                                                           c -r
                                                       •vi -3 ;

                                                       c  •
                                                                                                               r*\ O  -^
                                                                                                               r^ c  ••>
                                                                                                                                                                             -c
                                                                                                                                                                             O
                           ji o '
                           c.   »
                            J »•  O
                            AJ <3  T
                            ? -M  =T
                            O ••"  —
                            O   « —
                                                                     S  "VI  3
                                                                     =  — ^
                                                                     03-3-
3  >     —
                                                                                   J-
                                                                                   _  — —     >—3'     -« —
                                                                                       C-VJ     —  CO,     3-C
                                                                                       C  J-     .J-  =  3-     — ^
                                                                                                                             3«3—     3-C—    -yO—     TJOJ^
                                                                                                                             G   •—     S    •—    3    »—     ~   "»—
-T  T  —
J-  -T  3
3"  3  "^
-3"  "O =
                                          f  3
                                          —»  3
                                                        ^  O —
                                                                                                 .P —
                                                                                                 -T 3
                                                                     «—  'Xi »*     A O  «••     — 3
                                                                                                                                              — J-     ••  i
                                                                                                 V O  >
                                                                                                 f< o  a;
                                                                                                                                          X
                                                                                                                                                                      —  -U
                                                                                                                                                                      cr  X
—     X      "v  -r
                            O X  d
                            -r ^  c?
                            — !X  -c
                                                                         c j-     --
                                                        ^ a- \/>
                                                                     c;
                                                                      i
                                                                                   C  •« J
                                                                                   5  C ?
                                                                                                                                                            r -r     -z  - >r
                                                                                                      •        3
                                                                                                                                           IT J*
                                                                                                                                                         C 0-  "
                     •-     r  —
                                          .n />  c
                                          •^ p^  c
                                                                                                               <3 (?  X     =10.
                                                                                                               G   •        C    •
                                                                                                                                                ,  —    c-  r» ^*    o  -^ —
                                                                                    c  =  iri     r^  — o-
                                                                                   ex—     = r- ao
                                                                                                               e x
                                                                                                               o   «

                                                                                                               3

              ^  c  o     ir a
              ^3  \T aC     O O
                                          •f f  tr     f\i x  x

                                          >~ >  1?     1^  3  X
                                                                                   J"  3  C     3-1
                                                                                   "•  & O     J- "VI T

                                                                                   C  £ X*     O « «
                                                                                   C    •        O   •
                                                                                                                                                                       3  •v,
                                                                                                                                                                       C  !f
                                                                                                                                                                                    c  T .r
                                                                                                                                                                                    3-33-

                                          c
                                           r
                                                                                                                             i
                                                                                                                             ^1

-------
X





a





JC.
05     —      S
VJ
o
w
t^,
2
^.
_*
— *
^
*
r«
_,
—4
•r
C
u
c
-
^
^
to*
M
u
^
i"
j
a
^_
u
v.
'.r
**


c
4




c
a.
*-
X
X
j^
or.


^- —t/t
• o
C — i
< C 0
o o ^?
•^ O if
f\i •
» 0
c

IM — —
>C C —
*\j — *X
* o tr
• cr

O •** iT
— C J
C J^ U-l
o » «*•
• 0
c
1
fv. u~ cr
A. = tr
C.' • »—
• c
^

rv -r 3
/V •*• K-
-« — .f-
• ^
C

ir c r
_ «. IT
• • o
c

=C — If
X nu 5
if if »"
— C lf
C • *•
• C
c
»
a.
f%J t7 i/>
*
C. — X
•?• C ,C
=r c x
^V •
» o
o

•c — • —
_ c >*"
3- * ~VJ
"" 0 J-
« O

c — ir
o c =?
^ f\J Kt
c; rv :f
O • —
• ^
c
1
t/" O- =T
_ Z jf
Cr • v*
• -L.
O

or -? -y
>v = a-
• —
^

S^5
rv o if
""• c* ""
—

C Lf J"
tn rx =r
tr *•* **•
— C U"
^ « —
• c
c
i
a
Z- 0 «
'O O rf^
• 0
0
3 O <£,
iT c; tj
S, C \f
*- •
« -S
£•
I
•^ f^. —
^- •> "1
o *^  ^ lc'
AJ C ^
^c in ?•
C -C C
o •
• c
c
t
.X — X
SJ  ~ ^
O- C "„
^=,t/t
r*"""1
r
~- — IX
«; c <
A. C IT.
K1 • «•
* ^
C
I
^ i <.
-».—(/-
/••-•€
'V 0 0-
» "
c

•» a- *c
iv e ir
« c
—

^ H- /X
O *^ -C
cc » »••
C •*" tf
C' • •-
* c
c

c
o -c -^
O ^ iTl
o • —
* -O
c. c tr
c- •
» ^
c
t
^ c: ?
Cr O ff
— «— « iT>
— rw ^T.
* o
I
c rv *•
O r- —
C --, j-
o • —
• —
C-
1
tf- =r —
fV — tf
o • ^
• C:
O

«  ^ ^
m «- J*
c •
• —
c
1
^ j-v <•
C r» »
^ IP u"!
C -\' -0
• *
7
* •*• ^
c . 	
•C if r*
C; * VM
• c
c

c. c —
cr c j-
•^ * v
• —
VH

c c- r»
tr c IT
"• *" *"
c

£?£
if. e tf
• c
c

-x «. -^
J^ C *•*
fX C f*-
fKJ C J*
C • «•
• *
c

s
2 " Z
K* * If,
* •*• x
c o- »r
c »
» s-
—

j — ^~
^ ~ cr
'x — y
a, c, if
• o

c c ~-
T -r —
c c y
c. • —
• —
*»

" 7 *"
c ^r tf
c • *—
• V
^

!X = r*.
*V C If
* c
c
t
X c r*
*x c tr
• c
*
i
c ?x —
^ p*. «--
-c 
= ~ t
z — tr
— rv j-
c • —
• ~
c
1
^

-------
                                                              3-50
«(
>
i
             — 
-------
                               3-51
variables used  in  our analysis of chronic  morbidity is available  from




the authors.)




     For  example,  consider the simple  correlation between the average




daily  maximum  1-hour ozone  concentration  for  the  1U  day  reference




period  (03NR01)  and  the  average  sulfate  reading during  the  same  two




week  period  (S^NB01).    Table  3-6   indicates  that   the correlation




coefficient  is  0.119, that  this  is  significant  at better  than  the  1




percent  level  (p=.001),  and that this  coefficient was calculated  over




the 6,575 individuals for whom ozone and  sulfate  data were available




during  their  two-week reference period.  The corresponding correlation




coefficient  between  ozone  and  sulfur  dioxide   (S2NR01) is  -0.199,




calculated over 9,75U pairs of  readings.




     It  is  clear  from the length of  Tables 3-6 and 3-7 that we cannot




discuss  the  various  associations  between all the independent variables




used in this  report.  It is important  to  point  out that collinearity




does  not appear  to  be   a serious  problem  with  respect to  the  air




pollution  data  insofar  as  pairwise  comparisons  are   concerned.    The




highest  correlation  coefficient between  ozone  and any other pollutant




is 0.231  and  it is  below  0.20 in absolute  value for three of the  other




pollutants.     Ozone  is   highly  correlated  with  average  maximum



temperature  during  the   reference   period,  however,   the correlation




coefficient being 0.53.   This  is  to  be expected since  ozone is created




when  hydrocarbons and  nitrogen  dioxide interact  in   the presence  of




strong  sunlight,  a  condition  conducive to  high  temperatures  as  well.




Other correlation coefficients  are available  from Tables 3-6 and 3-7.

-------
                               3-52
     Because of the great number  of  different  models estimated in this




project,  we  did  not  print   out a  separate  matrix  of  correlation




coefficients for each and every regression.   All  models of adult acute




morbidity were  estimated  on subsets  of the  14,4^1  individuals used to




calculate the  matrix  in Table 3-6.   Similarly,  regressions explaining




children's morbidity  were based  on  subsets  of the  15»711  children in




the overall sample.   We have printed  out  the  correlation coefficients




for several  different subsets  of our  final  sample  and  the  resulting




data are very similar to those presented here.




     In  spite  of  our assertion that collinearity appears  not  to be a




serious  problem with  our  data, we have performed  diagnostic tests on a



number  of  the  regressions  estimated  below  using  sophisticated tests




which  permit  the  detection of aiulticollinearity when it  arises from




relationships  among  more than  two  regresscrs.   This  is  not possible




from an  examination  of  simple (bivariate)  correlation coefficients.  A.




description of  these diagnostic  techniques,  a list cf the  models to



which they were applied,  and  the  results  of  the tests are presented in



Section  u.5. below.








3.3  Estimation Techniques




     Given  the  different  measures   of  health  status  used   as  dependent




variables in  this  study, there are  many  statistical estimation  techniques




that could be  employed.   From this number, we have  tended in this analysis




to employ basic techniques  when confronted with several alternatives.  The




reasons  for this are  several.

-------
                               3-53
     First, as explained below, we  have  deliberately chosen in this report




to  explore  as many  hypotheses as  possible  in order  to  make use  of our




unique data base.   This broad approach  meant  that we could not explore  in




great detail  any one  finding,  thus  our choice of  fairly straightforward




techniques.   Second,  the methods  utilized in  this  report  tend to be less




computationally  burdensome  than some  of  the  highly sophisticated maximum




likelihood  estimation  (MLS)  methods  described  below.    All estimation




results  presented  in  this  report  were  computed  using  the  Statistical




Analysis System  (SAS).  While  SAS  has proven to be quite powerful in  terms




of  handling  the  large  data  sets  used in  this study,  it  is  also somewhat




limited in its MLS routines.




     We  turn  now  to  a discussion  of  the  various  estimation techniques




considered  in our  analysis,  including  those  actually utilized  and  those




that  seem,  potentially  fruitful for further  research within the  areas   of




concern of this study.




     The available  measures  of acute  health  status—RADs,  WLDs,  SLDs, and




BDDs—raise several  questions about  estimation methodology.   First, data



for RADs, etc.,  as collected  in the HIS are discrete,  assuming only one  of




fifteen  possible values  in  {0,1,...,1^}.   Days  where individuals  report




more  than  half  of  the day  as  limited  due  to some  illness  or injury are




coded as one  day,  with less than  or  equal to  half  the day coded as  zero.




In  the  present  study,  we treat  the   "days" measures of health  status   as




continuous on the interval [0,14], recognizing that there exist econometric




techniques to employ in the analysis of  such count data.  Others have  taken

-------
                               3-34
some initial  steps  in this latter  direction,  utilizing poisson regression


in the analysis of the effects of air pollution on work loss days.


     Consider now the  use  of  ordinary least squares  (CIS)  in the analysis


of  "days"  measures  of  health  status.    As  a brief  sketch,  consider  a


dependent  variable  vector  y   of  dimension  nxl,  a  regressor  matrix  X of


dimension nxk, a parameter vector  5 of  dimension kx1 and stochastic vector


of dimension nx1.   The econometric model can be written as:




                  y = X3 * £                                     (1)


and the OLS estimator of 3 is

                 'X      "It   t  rp
                 3  = CX-XT'X^y                                 (2)


This model is said to satisfy  full ideal conditions"' if


       1) X is ncnstochastio of rank k
-------
                               3-55
Halted dependent variable.  That is. for all possible values in the domain



X.  the  y  values  are  constrained  to  be  in  [0,14].   One  could  view  the



problem perfectly  analogously  by considering the measure of  health status



as  "percentage  of  the two-week period  restricted, in  bed,  at home  from



work, etc.," and view y as restricted to the range [0,1].



       Such range  restrictions  have definite implications  for statistical



estimation  of the  parmeter vector 3  in  ( 1 ) .   In particular,  for  finite X



and  3 .  we have a situation where the restriction on the  range of y implies



a restriction on the set of real (vector)  values that can be assumed by  £.



Specifically,  for  all  i. ^  must  be  in  f  -X.  3  .  1U-X.  3 ], i.e.,  for
                                               j.          i


continuous and differentiate f( ^ ) ,



                   14-X.3

                   -X.
                     i
where f( =:.) is the density of  z.
density of  the   £ _,  are technically inadmissable,  the  multivariate  normal

                                                           /\


being the most  obvious.   Obtaining an unbiased estimator  2  of  3  is  still




possible with OLS as long as E( e ) = 0.  but the statistical tests  based  on




the  assumption  of £ being distributed  multivariate normal are no  longer




strictly appropriate.  However, if the full  ideal  conditions  are met,  save




for the normality assumption, the usual  test  procedures  based on normality




can be asymptotically justified so long as Z(£ )=0. Var(£ ) is finite,  X  is


                             Tl                                •>

uniformly bounded, and lim (X X)/n is finite and ncnsingular.

-------
                               3-56
       Cur  estimation  of  the  "days"  models  res;:  on  several  assumptions.


First, the  large  number  of observations  in  the  ace-els estimated  allows  us


to invoke  the  asymptotic properties of  the  estimators.   Second,  it  cannot


be ruled  out  ex  ante that  Z( £ )=0,  although  given  that  there  exists  a


considerable  mass  point  at   £.=-X,.3    (i.e.  no  days   resorted),   actual


specification  of  the density  of £  is  not  attempted  here.   Finally,  and


quite practically, computational cos's  are mucn lower for CLS  estimation  as


compared with MLZ methods, which require the further step of  specification


of the fora of the density of t  ,  We new concern ourselves witn discussion


of other  methods  appropriate  to alternative specifications  of (''  above.


Some of these are in fact utilized in our analysis , others are  presented  to


stimulate  thinking about  farther research bevcnd the bounds cf this  stunv.


For  new,  we merely  state that  OL3  is  widely  used In  the  present  study:


'vf.ere used, one must re aware of the limitations  described above.


       An approach considered frequently in econometric analysis of limited


dependent  variables  is  the use  of latent variables as  censoring  agents.'

                                                                *
That  is   to   say,   we   consider  some  continuous  variable   y    teat   i.3


meaningf-ally modeled as


                  7* = 33 - z                                    (3;


where  £  i3 well-behaved in the classical  sense.  For simplicity, we will


assume  that  £ is  distributed  independent,   multivariate  normal,  although


this  is  an  inessential  restriction  for  many  results   presented.    The


Important  thing  to  note  is  that there  are  no  restrictions  imposed on the


domain of   £   ; we  assume only that f ( e )  obeys the usual restrictions  on


probability density functions.

-------
                               '3-57
       However, in the limited  dependent  variable model, we  do  not  observe
 *                                    *
y ,  or  at  least we  do  not observe  y  over its  entire range.  Rather,  we
                                                 *
observe  some  variable  y  that  is  related  to  y   by  the  following  quite
general relationship:
                  y  = y* If CT < -/ <  cu,
                                 *
                        = £-,_ if y  <_ c,
                        = G   c  ,
                           u -  J  —   u'
where c, ^-» ,c  )  and  c,  ^ (G, ,+ =°  ] are constants.   The problem of  the
analyst  is  to  estimate the parameter  vector 3  in  (3)  in a  consistent  and

efficient manner.
       It  is  Glear  that  such  a  model  is   germain  to   estimating  the
determinants of  the  "days" measures  of  individuals'  health status.   Note
that the presentation of  the  model as a  latent variable  model  differs  from
our  discussion above  regarding  CLS   estimation.    Insofar  as  ^he ''days"
measures are concerned, the definition of y  is not  important.

       ?or  present   purposes,  y   is  something like  "poorness of  health,"
where,  loosely speaking,  larger values of y  imply increasing  probability
of  additional  HADs,  WLDs, 3DDs, or SLDs.  The bounds  on the  observability
of  y   are  those imposed  by the nature of the "days"  data, i.e. no  fewer
than zero and  no greater than fourteen "days" can be reported  in a  two-week
period.
       If  one  decides  to  take  a latent variable  approach to estimating

"days"  equations,  there  are several routes  that  can be taken.    Some
analysts have  suggested that because of the  considerable  mass  point  at  zero
days,  and  because   there  are  so  few individuals  who  report  upwards  of

-------
                               3-53
fourteen days of restriction, a  tobit model  would  be  an appropriate one to



estimate.   In the  classical  tobit model, c   is irrelevant  (c   =  +  » in
                                             u                  u


terms of  the above presentation)  and  c,  is set at  zero.   The  likelihood



function for the tobit model is



              L=  U ~-1f((y.-x,  3 )/c )   II?(-x.  3/J)           (4)

                   si        x  *         s2     x

where S   and S,  are  the  noncensored and  censored  sets of  observations.



respectively, and f(«) and  ?(•)  are  the  density  and  distribution functions



of a  standard normal  variate (we here assume that z  in (3)  is distributed



independent  normally).   The parameter vector (  3T ~ )  can be  estimated by



means of maximum likelihood :r by an  iterative  least  squares  procedure.'



       Recognition  of  the  limitation at  fourteen  days,  however,  suggests



that  a two-limit  probit  estimator be used.1    Here, c,  and  c, are  finite,
                                                       -t.       ^


and for purposes of exposition,  it is assumed that  there exist observations



at  c.  and  c  and  in-between,  as is  the case  in the  "davs" data.   ~''-^e
             u


likelihood function for this model can be written as



        L = ,~  ?((c,-x. 3 )/  J ) ~FCx. 1 -c )/  ~)-~^~1f((y.-x. 3 )/ c]  (^
            &1      xi       S2     i    u    s3     'i  i


3., s_,  and S, are  the sets  of observations  at the lower and uoner  sounds,
      2'       3


and in-between,  respectively.  Estimation of  3  is again performed by



maximum  likelihood.   Again,  normality  and  independence  cf  the  error



structure in (3) are assumed.



       As  mentioned earlier,  yet another  approach  would  be an  explicit



recognition of the  discrete  nature of the data,  with  estimation  under the



assumption of a poisson distribution for the days.  However,  the mean  of a



poisson-distributed  variate is  equal to  the  variance, thus imposing  a



sometimes 'unrealistic moment structure on the empirical distribution of the

-------
variates.   Furthermore,  the  censoring at fourteen  days  again implies that




the  strict  poisson  model  is  not  appropriate.   However,  the  dearth  of




observations at fourteen would most likely  imply that a poisson model with




explicit  recognition  of  the  fourteen  day  limit  would  yield  parameter




estimates  hardly  different  than  one  in  which  the  censoring  limit  was




imposed in the  likelihood  function.   Again,  estimation  is accomplished by




methods of maximum likelihood,  or by a weighted least squares approach.




       To  summarize,   the   "days"  measures   of   health   status  present




interesting and  challenging estimation  problems.    Whether  it is  used or




not, the  latent  variable approach has  been the methodology  most studied,




and, therefore, the one on which the most information is available.




       However, many  interesting questions  concerning  individuals' health




are  of  a  different type.   Does  an individual have  a  certain condition or




set of conditions?  Was an individual restricted in activity to some extent




during  a  particular  period?    Because  measurement  of health  status  is




definitionally  difficult,  many  of  the  most  illuminating questions  about




health are necessarily  framed  in such binary or  discrete  terms.   That is,




attention is confined  to the presence  or absence of a particular condition




or illness.    (This  is not without its  own  problems, since two individuals



both reporting  chronic  sinusitis,  for instance,  might differ substantially




in the severity of those conditions.)   Fortunately, much attention has been




devoted recently to analysis of discrete outcomes (by discrete outcomes, we




mean situations where an outcome occurs or  it  does  not—as  in the case of




an individual who either has a respiratory disease or does not).    Here we




present the basics  of the  analysis  of discrete outcomes  which is  used in

-------
                               3-60
much of  our analysis  below.   We  restrict  our attention  -o  binary choice




since that is the approach employed in our study.




       The  basic  idea  behind  the  binary  outcome  model  is  that   the




probability of  an outcome (an illness,  for  instance)  depends functionally




on a vector of variables or attributes characterizing an individual and  his




environment, broadly  defined.   Much  of  the analysis  of  discrete outcomes




has  been cast  in the framework  of  individual  utility  maximization:   an




individual chooses the outcome which  is  higher in his preference ordering.




For  some analysis,  however,  the  notion  of  deliberate   choice  is  not a




particularly appealing one.  Although some  might contend  that all outcomes




are to some extent ''chosen,1' it stretches one's normal understanding of  the




meaning  of  "choice"  to  think  of  outcomes  like  chronic  cardiovascular  or




respiratory disease  as being "chosen."   For  present  purposes,  it is much




nore straightforward  to  view  outcomes  from  an  ex  oost  vantage  point  and




consider the characteristics of an individual and his environment that




might  be related to  the observed   (health)  outcomes,  and  how  they  are




related.



       In this  study  binary outcomes  have been  examined  in both the acute




and  chronic  analyses   undertaken.   For  instance, acute health  outcomes  of




the  binary  nature that  have  been studied  are the  presence  or  absence  of




restricted-activity,  work or  school  loss,  and  bed-disability days.    Our




analysis  of  chronic   morbidity  is based  on  the  presence  or  absence  of




chronic respiratory, cardiovascular,  and several  other conditions.




       We describe  the  probability  that individual  i has outcome  y..   (at




least one bed  disability day,  for example)  to be a  function of a vector of

-------
                               3-61
characteristics X. and a parameter vector  3 .  That is




                    Pr(y  = illness) = F(X± 3)                    (5)




     For convenience,  we define y. = i  if the binary  outcome  is a "yes"  or




illness and  y^=o if  the outcome  is  "no"  or  health, and  assume that all




individuals are characterized by either an  illness or health, but not  both.




     The fact  that we  are examining  probabilities  of  outcomes  implies  a




special restriction on  the range of the  function  F,  i.e.,  FCX..3 ) £[0,1].




There is an  obvious  class of functional  forms  that  definitionally imposes




such a restriction on  the  functional  range, however,  and this  class is the




distribution functions  corresponding to  continuous  probability  densities.




We assume, therefore, that X,.3 ±s continuous on (- =°  ,->- °° ) and examine the




nature of  the  functions FCX..3  ) such  that  F:B—>[0,1].   The main question




is: What is the fora of F most appropriate  for analysis, or, equivalently,




what is the appropriate form of the density f(3)=dF(2)/dz of the  'underlying




scalar-valued "indicator" variable z.=X.,3   ?




     Although it  does  not necessarily satisfy  (5),  the  linear probability




model has  been  used frequently  in many  different area  of  analysis.   This




model assumes the form




     ?r(yi=D = X.,3,                                              (6)




i.e. F(Xj,   )=X13.  In order to satisfy (5), (5) is recast as




     ?r(yi=D = o, X13  [.» >0]




              = Xj_ 5, Xl3£(o,1)                                   (7)




              = 1, X13 S[1,+ ao]




     There are several detailed descriptions available of the properties  of




the linear probability model, the major drawback of which is that predicted

-------
                                3-62
probabilities,  I.  3,  Gan be  outside  of [0,1].     That is,  an individual



might  be predicted to have  greater than a  100 percent  chance of having an



illness.   A corollary problem is  that individuals who  are  predicted to be



certain  of having  a  particular condition  or  illness  (p=1), may  not have



that condition  at  all.



     This   aside,  the   linear  probability  model  is   straightforward  to



estimate.   One  writes




     7±  --  X1 3  -  £,,                                             (3)



where  y_.  is  the  nxi  vector  of  binary  (0,1)  outcomes.   Clearly,   z.  is



heteroskedastic.    Furthermore, it can  assume  only  one  of two  possible



values,  -X. 3   Or  (1-X. 3 }.    For  2(£ .)=0. one  must  have Pr(  =.=-X.3  ) =
           i          —                _                          i    i.


(1-X. 3) and ?r( £.. = 1-X,. 3}=x,.3.    Given  this  restriction and the face chat



the  variance  of a binomial  variate with parameter ^  equals - (1-  3  ],  one



has  Var(  ;. ]  - X.. - (1-X_. -  ". .   Thus.  -  ir.  (^',  or (3/   can  be estimated by



GL3.



   3 - ,'TCT ~-1vv-1v' --"!-,                                         -^
   irvA^A;A^/                                         \^)



where  ~ =diag("ar( £ _. ) )=diag(Xn.   3 'l-X.  3  )).   Since -  is  unlikely to be



known, a  feasible GL3  ^FGLS) estimator  would  use consistent estimates of



7ar( € _. )  (using, perhaos .  3     to predict  X_.3  )  and  have
       j.               *      wLo             j.

   =:     ~n   *   . m ^  -.

   3 = (X1^ -'xrVf. ~1y                                         (10)



where  ~  =diag(X   3    (1-X.  3   )) for X.  3     £ (0,1).  It can  be shown
                1   ULo     j.   JLo         -   JLJLJ

            T  ^ _ 1                                     T'N— 1
that  plimCX*  -- ~  X)/n   finite nonsingular  and   plimCX  -    ^  )/n=0  are


                                               =                     T    •*
sufficient conditions  for the  consistency  of  i,   and that if piim(Xi ( " ~ '_



J~')X)/n=0  and  plimCX  ( H" -  H " ') £ )/n~°=0  then  the  FGLS estimator has  the



same  asymptotic  distribution  as   the GLS  estimator  and,  therefore,  is

-------
                               3-53
efficient.      One  muse   be  careful  when  using  FGLS.   however,   as  the



extremely high  weights on  observations  with  predicted probabilities near



unity  or  near zero  can cause  considerably  more  variability  in parameter



estimation than would result from unweighted OLS estimates.  This situation



is particularly problematic in the small sample case.



     The  main  appeal  of  the  linear  probability  model  rests   in  its



computational simplicity and low cost, in either the OL3 or the FGLS  cases.



We utilized  linear probability analysis  in  the preliminary stages  of our



analysis,  but because  of  the  limitations  discussed  above,   it was  not



employed in estimation of the models presented in Chapter  4.



     The two most  popular  and  most  utilized  methods for analysis of  binary



outcomes  are  the  binary_ probit and  binary  logit  models.  In  the   binary



orobit model, one has
     ?r(y.=i) =  /f(z)dz = F(X. S/ r),
where  z~N(0,1).    The  parameter vector  1  is  estimable up  to  a  scalar



multiple  ("/ j )3 without prior  knowledge  of ~  .   In the logit model, one




has




     ?r(y, = D =  (1 + ex?(-X, 3))"1 = F(X, 3).                   (12)
         -L                  J.            -L


On the  assumption  that  observations are independent  draws  from a binomial




distribution, the likelihood function for  these  two models,  can be written



as




     LCy. ) =  .1 F.yi  Hd-F. )°-yi)                              (13)

        1    Si   i    S2


where Fn. =F((X_. 3  )/j ) for  binary  probit  and F,. =F(Xn. 3 )  for binary logit,

-------
                               3-6U
and  where  S^  and  S-,  are the  sets of  observations  of  ill  and  healthly



Individuals (y.=i and y.=0),  respectively.  Estimation of both models is by



maximum  likelihood,  maximizing L  in  (13)  with  respect  to  the parameter



vector.  Logit  modeling  has  tended to dominate  probit in applied analysis



because  of  logit's  closed  form expression  for  F.    In  probit, for  each



observation at  each iteration  there  is  call to the normal  distribution



function at the  value at  which  the  integral  in (11)  must be evaluated.  We



have  performed  the  bulk  of  the  binary  discrete outcome analysis  in the



present  study  in  a  logit  framework,  using SAS's  PROC  LOGIST  maximum



likelihood routine.   In  many cases  the  parameter estimates  obtained  from



MLS  of  the  logit  and   probit models   are   quite   comparable,  up  to  a



multiplicative  constant,  and  there  exist  formulae  for the  approximate


                                                        1 a
conversion between logit  and probit parameter estimates.


     Both the logit  and probit  models  can be  estimated by procedures which



allow  for  iterative  reweighting of  the  observations in  a  nonlinear least



squares framework  (SAS's  ?POG MLIM  and 3MD?'s ?3S and PAR,  for example).1"^



Such  estimation is  by itsratively reweighted  nonlinear 7GLS  (IHNFGLS),



based  on  explicit  recognition of the  (assumed)  independent  binomial trial



structure of the y, vector.   One has



     y± = F(zi) +  £ ±                                            (14)



where the properties of  E; .  are basically those discussed above in the case


of  linear  probability models.   One  iteratively  searches  for  the  vector


that solves



     min (y - F)T 2 ~1(y - F)                                    (15)



where   ~    is  a  diagonal iteratively updated weight matrix  with  typical

-------
                               3-65
element [(?(X.2  . )(1_F(X.3  O))~11» where  3   is the  orevious  iteration's
             1 — I       1—1                — !



estimate of  3.  It  can be shown  that  IRNFGLS applied to  (1H)  is maximum




likelihood given some regularity conditions.  Mote that since many packages




have functions  or  calls to routines that  efficiently  calculate  cumulative




standard  normal integrals  (e.g.  the   PBOBNORM  function  in SAS),  probit




estimation is quite feasible.




     To summarize,  since  many  health  outcomes of  interest are  binary in




nature,  one  is confronted  with  a  discrete-outcome   model.    The  above




discussion has identified a set  of  the  estimation techniques available for




such analysis.   Much  of the potentially interesting future research might




be conducted in the framework  of the multi-outcome extensions of  the binary




outcome models examined here.

-------
                               3-66
                          Footnotes to Chapter 3



      See Atkinson, Crocker, and Murdock [1983].


     ^CHS [1981], pp. 13-20.


      There are  two  very comprehensive  summaries  of the  effects  of ozone


and  other photochemical  oxidant  pollution  on  health.    These  are  U.S.


Environmental Protection Agency [1978] and the National Academy of Sciences


[1977l.

     4
      See Goldsmith and Landaw [1968], for instance.


     "'See Hausinan, Ostro and Wise [1983].

     5
      See Schmidt [1976], p. 2.


     'See Schmidt [1976], pp. 56-50.

     Q
     "See the surveys by Amemiya [1931] and Maddala [1983".

     q
     'See Fair [19"7].

    1 o
     '"See ?ose~t and :ielson [19'7r].


      See the Amemiya [*981] and Maddala [1983] surveys.


     "See Dcmenich and McFadden [1975].


     ^See Schmidt [1976].

    "• U
      See Amemiya [1981], pp. 1487-1488.


     'See Jennrich and Moore [1975].

-------
                               3-67
                          References to Chapter 3
Amemiya, Takeshi.   "Qualitative  Response Models:   A  Survey,"   J.  Scon.
     Lit.,  19(1981), pp. ^85-496.

Atkinson,  Scott,   Thomas   Crocker  and  Robert  Murdock,  "Have  Priors  in
     Aggregate Air  Pollution Epidemiology  Dictated Posteriors?",  working
     paper from the Department of  Economics,  University of Wyoming,  1983.

Domencich,   Thomas   and  Daniel McFadden,  Urban  Travel Demand  (Amsterdam:
     North Holland), 1975.

Fair,   Ray,   "A   Note  on   the   Computation  of   the  Tobit   Estimator,"
     Econometrica,  15(1977), pp.  1723-27.

Goldsmith,   John  and  Stephen Landaw, "Carbon  Monoxide and Human  Health,"
     Science, 162 (1968),  pp. 1352-1359.

Hausman, Jerry, Bart Ostro,  and David Wise,  "Air Pollution and Lost Work,"
     paper  presented  at  the NBER  Conference  on  Productivity in  Health,
     Stanford, California, August  18-19, 1983.

Jennrich, R.I. and  R.H. Moore,  ''Maximum Likelihood Estimation by Means  of
     Nonlinear  Least   Squares,"   American  Statistical Association,   1975
     Proceedings of_ the Statistical Computing Section,  pp. 57-65.

Maddala, G.3., Limit ed-Deper.dent and Qualitative Variables  in_ Econometrics
     (Cambridge:   Cambridge University  Press),  1983.

National  Academy  of   Sciences,  Ozone  and  Other  Photochemical   Oxidants
     (Washington,  D.C.:  National  Academy  Press),1977.

Rosett,  R.N.  and  F.D.   Nelson,  "Estimation   of   the   Two-Limit   Probit
     Regression Model," Eccnometrica, 43(1975), pp.  141-116.

Schmidt,   Peter,    Econometrics   (New   York:     Marcel   Dekker),   1976.

U.S. Environmental  Protection Agency,   Air Quality  Criteria for Ozone  and
     Other Photochemical Oxidants, Publication no.  EPA-600/3-73-OOU,  April
     1978.

-------
                                 A 3-1
                         Appendix to Chapter 3


                  DEFINITIONS OF AND PROCEDURES USED
                    TO CREATE MEASURES OF MORBIDITY

The present study, measures of health status, can be separated into two

distinct  groups.    First  are  the  "all-causes"   measures  of  health.

Second are  more  specific measures, dealing  primarily  with  respiratory

conditions, but also  to  a  lesser  degree  with cardiovascular illnesses.

Measures   in   the  second   group  are  further   disaggregated  where

appropriate into acute and chronic categories.



In  the  following  descriptions,  references  are  made  to  definitions

presented  in  Chapter 2  and Appendix  A  to  be  as  precise  as  possible

about the definitions of the several measures of health status.



Within  the all-causes  classification,  six  measures  of health status

have been utilized.  They are:



1) RADMIMOR -  This is a measure  of  total  two-week restricted-activity

days when  such restrictions are  of  a minor nature.   This  variable is

formed as follows:

     For adults; RADMINOB = TOTRADZW if BSDDIS2W = 0 and WXSLLD2W = 0

                          = n/a if BEDDIS2W > 0 or WKSLLD2W > 0.
     For children:
                 RADMINOR = TOTRAD2W if 3EDDIS2W = 0 and WKSLLD2W = 0,

                            if WKSLLDID = 2

                          r TOTRAD2W if BEDDIS2W = 0, if WKSLLDID = 1

-------
                                 A3-2




                          = n/a if 3SDDIS2W > 0 or WXSLLD2W > 0,




                            if WKSLLDID = 2




                          = n/a if BSDDIS2W > 0, if WKSLLDID = 1.




The reasons for conditioning the definition of RADMIMOH for children on




the  value  of WKSLLDID  is,  of  course,  that by  definition  children of




pre-school age will report zero for WKSLLD2W.









2) WKSLLD2W - As defined in Chapter 2, subject to




     For adults; WKSLLD2W is defined only for CURACT2W = 1 and




                            CURACT2W = 2, n/a otherwise.




     For children;




                 WKSLLD2W is defined for WKSLLDID = 2 if PROCQUAR = '




                 or PROCQUAR = ^ or PROCQUAR = 2 and WEEK < 10 or




                 PRCCQUAR = 2 and WEEK > 9: and n/a otherwise.




"or  adults,  the  CURACT2W cutoffs were established  so  to consider only




those  individuals  for  whom work  loss  is  a  meaningful concept.   The




rationale  for  the  cutoffs  for  children  is   to  proxy  as  nearly  as




possible  the boundaries  of a  representative  school year.   Obviously




only those  children  of school age  (WKSLLDID  =  2)  qualify for analysis




in a school-loss days model.









3)  BEDDIS2W - As  defined in  Chapter  2.   This variable  serves  as an




aggregate measure  of  restrictions  in activity of a severity sufficient




to warrant bed-disability.









U) RADPRESM  - This variable is derived from RADMINOR as follows:

-------
                                 A3-3




                 RADPRESM = 1  iff HADMINOR > 0




                 RADPRESM = 0  iff HADMINOR = 0.




This binary  measure of minor  restricted-activity days is  of  interest




when questions of the "whether or not" or "have, don't have" are posed.









5) BEDPRES - This variable is  derived from 3EDDIS2W as follows:




                 3EDPRES = 1  iff 3EDDIS2W > 0




                 3EDPRES = 0  iff 3EDDIS2W = 0.









6) WSLPRSS  - This  variable is  derived from WXSLLD2W,  subject to both




the adult and child restrictions described above, as follows:




                 WSLPRES = 1  iff WKSLLD2W > 0




                 WSLPRSS = 0  iff WKSLLD2W =0.









Insofar  as  respiratory conditions  are concerned,  seven  measures have




been  utilised,  six  acute and  one   chronic.    The  following  table




illustrates  those measures in  the Health  Interview  Survey that we have




classified as  respiratory conditions.   The three-digit condition codes




listed  are  values assumed by  the  variables DGRCD101, DGRCD:02,  ...   ,




DGRCD11U defined in Chapter 2 and Appendix A.

-------
Table .4.3-1.  Three-digit Recedes—Diseases of the Respiratory System
   14U.  Common Cold *




   145.  Other Acute Upper Respiratory Infections *




   1^6.  Acute Bronchitis *




   1^7.  Influenza with Digestive Manifestations *




   "^8.  Other Forms of Influenza *




   1^9.  Viral Pneumonia *
   "i5".  Chronic Bronchitis (nonallargic'!




   *^2.  Zaphysema




   153.  Asthma (with or without hay fever) (allergic)




   15^.  Hypertrophy of Tonsils and Adenoids,  Chronic




   155.  Chronic Pharyngitis, Nasopharyngitis , and Laryngitis, MEC




   156.  Chronic Sinusitis, NEC




   157.  Hay "ever, and Upper Respiratory Allergy, without Asthma




   158.  Other Diseases of Upper Respiratory Tract (nonallergic)




   159.  Other Diseases of Respiratory System, NEC




   228.  Certain Respiratory Symptoms *
* Can be coded as acute if onset is less than three months prior.

-------
                                 A3-5
The respiratory condition measures are:







7) TRADRSP  -  This  variable is defined as  the  total number of two-week



restricted-activity days  due  to  respiratory conditions.  Specifically,



for i indexing over {01, 02,  ...   , 14},



                 TRADRSP = Max.. {  RAD2W., i  DGRCD1., in Table A3-1 above  }



That   is,   TRADES?   is   the   maximum   of   the   set  of   two-week



restricted-activity days given that such days were due to a respiratory



condition.  If no respiratory-related two-week restricted-activity days



were reported, then THADRSP = 0.







3) TWSLRSP  -  This  variable is defined as  the  total number of two-week



work- or school-loss days due to respiratory conditions.  Specifically,



again for i indexing over {01, 02, ... , 1 ^ },



     For Adults: WSLRES? = Max, {  WSL2W, 1  DGRCD1,. in Table A3-1 above  }



                 if CURACT2W = 1  or CURACT2W = 2, and n/a otherwise.



     For Children;



                 WSLRESP = Max.{  WSL2W. i  DGRCDI^ in Table A3-1 above  }
                              11         0.


                 if WKSLLDID  = 2, and n/a otherwise.



The  maximum  here,  of  course,   can  be   zero  for  those  individuals



reporting no respiratory-related work- or school-loss days.







9) T3EDRS?  -  This  variable is defined as  the  total number of two-week



bed-disability  days  due  to respiratory conditions.   Specifically, with



i again  taking on  values 01,  ...   , 1^,



                 BEDRESP = Max, {  BED2W.. i  DGRCD1,. in Table A3-1 above  }

-------
                                 A3-6




Again, no respiratory-related bed-disability days implies BEDRESP = 0.









10) RADRSPPS -  This  is a binary variable,  constructed  from TRADRS? as




follows:




                 RADRSPPS = 1 iff TRADRS? > 0




                 RADRSPPS = 0 iff TRADRS? = 0.









11)  WSLRSPPS  -  This  is  a  binary  variable,  derived  from  WSLRSSP  as




follows:




                 WSLRSPPS = 1 iff WSLRESP > 0




                 WSLRSPPS = 0 iff WSLRES? = 0.









12) 3EDRSPPS -  This  is a binary variable,  constructed  from 3EDRES? as




follows:




                 3EDRS?DS = 1 iff 3SDRES? > 0




                 3EDRSPPS = 0 iff 3SDRES? = 0.









13)  PRSCHRSP   - This  is  a measure  of  chronic respiratory  illness.




3ecause  of  the nature  of such  chronic  conditions,  a  discrete  binary




measure was deemed appropriate.  PRSCHRS? is defined as




                 PRSCHRS? = 1 iff, for any i in  { 01, ... ,  14 },




                      DGRCD1.. is in Table A3-1 above and CRNACT,.  = 1




                 PRSCHRSP = 0 otherwise.




Thus  PRSCHRS?  assumes a  value  of  one when  the  individual  reports any




respiratory condition  listed in Table A3-1  defined  by the standards of




the HIS *o be chronic, and zero otherwise.

-------
                                 A3-7




Preliminary explorations into the possible effects  of  air  pollution on




cardiovascular illness have been undertaken in this stuny.   For present




purposes, cardiovascular illnesses  were classified as  those  values of




DGRCD101 through DGRCD114 listed in  Table A3-2 below.
Table A3-2.  Three-digit Recedes—Selected Diseases of the Circulatory




             System
128.  Chronic Rheumatic Heart Diseases




129.  Ischemia Heart Diseases (with Hypertension, Any Type) (with




      Arteriosclerosis)




130.  Heart Trouble, NOS, or Ill-defined




131.  Other Forms of Heart Disease, NEC




"32.  Hypertensive Heart Disease, NEC (Non-malignant)




133.  Hypertensive Disease, NEC




134.  Cerebrovascular Disease (with Non-malignant Hypertension) (with




      Arteriosclerosis)




135.  Arteriosclerosis, NEC




136.  Other Diseases of the Arteries




141.  Poor Circulation, NOS




143.  Other Diseases of Circulatory System, NEC

-------
                                 A3-8
The  single measure  of  cardiovascular conditions  used  was  PHSCHCRD,




defined as:




                 PRSCKCRD = 1  iff, for any i in { 01, ... , 1U},




                      DGRCD-L  is in Table A3-2 above and CRNACT., = 1




                 PRSCHCRD = 0 otherwise.




Thus, PRSCHCRD assumes  a value of one  when  the  individual reports any




cardiovascular  condition  listed   in  Table  A3-2  and  defined   by  the




standards of the HIS to be chronic, and zero otherwise.

-------
                                 Chapter 4









                          RESULTS AND DISCUSSION









     In  this  chapter  we  report  our  findings  about   acute  and  chronic




morbidity and their relationship to ozone and oth.'r air pollutants.  First,




some introductory remarks are in order.




     In  Chapter  1,  several  problems  inherent  in  epidemiologies!  studies




were discussed.  Other problems arise in attempts to estimate dose-response




relationships  via  statistical  regression.   First, this  approach recuires




one to  specify a functional form  for the dose-response  relationship,  yet




there is  very little theoretical  guidance  as to its likely  nature.   Seme




studies   use   a   simple  linear   form,   others  rely   on   a  logarithmic




specification, while many  recent studies have  employed  an  S-shaped probit




or logistic  functional  form.  There  are other  possible  choices, as  well,




with examples of each in the literature.




     Nor is there consensus about the "target populations" to be studied or




about  the  sorts  of  adverse  health  effects  on  which  to  concentrate




attention.   One  can  concentrate  on  acute  or  chronic  morbidity;  examine




respiratory  as  opposed  to  other effects;  focus  on the  very young or  the

-------
very old, or the general  population;  concentrate  on the healthy as opposed




to the infirm;  and so on.




     In  addition,  there  is  disagreement about  the  explanatory  variables




that should  be  included, or how  they should be  expressed,  as  for example




whether   air   pollutants   should   be  measured  by   peak  or   average




concentrations.   Some  studies  use a  single air  pollutant  to characterize




exposure   to   all   pollutants,   some   include   few   socioeconoraic   or




meteorological measures, some fail or are unable to control for the smoking




habits of individuals.   In most studies,  as in the present one, the number




of  control  variables  is  determined  as  much  by  the  availability  of




information  as  by  the  conscious  choice  of the  researcher  to  include  or




exclude certain measures.




     Together,    these   factors—the   range   of   possible   and   plausible




functional forms, types  of  health effects to  be  studied,  choice of target




populations,  and   variety   of  explanatory   variables  to  consider  for




inclusion—have  important implications first for the design  or  conduct  of




research  and second for the  presentation  and interpretation  of  results.




Concerning  the   former,   they  suggest  that,  where  possible,  one  should




experiment with a  variety  of  functional forms,  types  of  health  effects,




target populations and control  measures.




     That is the approach taken in this report.  Where  a  choice  had to  be




made, we  elected to  explore  a  broad  range  of possible associations between




air pollution and acute  and  chronic  disease using fairly standard ordinary




least squares and logit techniques.   Because of the several 'unique features




of our data, it seemed advisable to take this route rather  than confine our

-------
attention to a few relationships which would then be subjected to a battery




of sophisticated  tests.   Although the approach  in  this  report is designed




to  help  identify  "targets   of  opportunity"  for  further  research,  the




relationships  we  estimate do  not  go  unexanined.   The results  to  follow




include  a great  many  sensitivity  analyses,  and  sophisticated  regression




diagnostic procedures have been  applied  to  several  of the most interesting




or important findings.




     With respect to the presentation of results, such an approach means it




may  be  virtually impossible  to present in a  report every model  that was




estimated.   In a  study like the  present one the number of separate models




explored  often extends into  the hundreds,  thus  forcing a  decision  about:




which  results  to present.   In  making this decision,  it is  important to




convey  the  full  range  of  work done and results  obtained.   This  is  to be




contrasted with an approach in which only the "best'1 results are displayed,




that  is. those  which  enable  one  to  either   reject or  accept  the  null




hypothesis   of  no   relationship  between   air  pollution  ana  morbidity.




Selective reporting can give a misleading picture of this relationship.




     The  presentation  of  results  in  this  report  is designed  to minimize




this problem.   The  results presented here  represent  fairly well the range




of  functional  forms, categories of health  effects,  target  populations and




combinations  of  explanatory variables  explored  in the  study.   Generally,




within  a particular area of inquiry—acute respiratory  disease  in adults,




for  instance—several  versions  of  a  "basic"  model  are presented.   These




versions may differ with respect to the number of air pollutants besides

-------
ozone which are analyzed, the way  these  pollutants  are measured, and other


dimensions.


     Our approach could be termed conservative in several respects.  First,


time  and   financial   constraints   made  it  impossible   to   analyze  the


sensitivity  of  every  finding.    We  elected  to  do  much  more  sensitivity


analysis  for  those models  in which  the  null hypothesis  of  no  effect  of


ozone on  acute  or  chronic  morbidity was  rejected  at  standard  confidence


levels.    The  goal  was  to  explore  the  observed  effect  for  statistical


robustness.  In  some  models  where no statistically  significant  effect was


observed, little  or  no  additional analysis  was done.  In certain  cases,


however, sensitivity tests were performed even if no significant effect was


observed.


     Another respect  in  which  our approach  might  be  termed  conservative


concerns hypothesis testing.  Throughout  the  results,  two-tailed tests are


used to  determine  the significance of the coefficients  on ozone and other


explanatory  variables.   Using  two-tailed  tests,   t-ratios  on  estimated


coefficients must be  >_ 1.96 for  the  coefficient  to  be significant at the 5


percent  level and  > 1.65 for significance at the 10 percent level.


     This   approach   might    be    termed   conservative    because   some


epidemiological studies  have  used  one-tailed tests  on  the  grounds  that  if

                                                                          1
air  pollution has  any effect  on health,  the  effect  will only be adverse.1


When using a one-tailed  test, the  t-ratio on an estimated  coefficient need


only be  1.55 for the coefficient to  be  significant  at  the 5  percent level


and  1.23 for significance  at  the  10 percent  level.   The more  rigorous


requirement associated with the two-tailed test is used here.

-------
     The parameter  estimates  and  t-statistics presentee below are  computed

using the usual formulae:

               -1                 ,                   -1
     5   = (XTX)  XTY    and  t. - £,SQRT roiAG(S2(XTX)   ),l.
                              ~    -U     *"               —

For each OLS regression, we  present  at the bottom of the column  of  results

the  unconnected  ?.-, the  "-statistic  testing the hypothesis  H  :  3   =  Q

(where 3 ,  is the parameter vector except for the estimated intercept),  and

the number of observations in the  estimated equation.

     For logit models,  we  present the maximum  likelihood  estimate  of  each
           ^                                                   ••
parameter (£„.) and  the  estimated  t-statistic.  The latter is  3.  divided  by

the  square root  of  the corresponding  diagonal  element  of  I~ ' evaluated  at

the  maximum  likelihood parameter  estimates  (I  ±3  the information matrix).

J"  the  bottom of  the  columns  presenting  logit  results are  found  the  "
-------
                                    4-6
morbidity used in our study.  Unfortunately,  the  HIS makes it difficult to

identify these minor  restrictions  because any work  loss  :>r  bed disability

day is also coded as a restricted activity day.  Tc be sure we were working

with  truly  minor restrictions,  we  confined  this  analysis  to  only  those

individuals  reporting  RADs  who  did  not  also  have  work  loss  or  bed

disability days.  The sample  thus  excludes  those  who reported more serious

acute illnesses during the  reference  period.   The acute morbidity analysed

here was that due to all causes, not just respiratory.

     The  first  regressions,  presented  in Table  ii-1, used  ordinary  least

squares to relate MRADs  to  ozone (as  measured by  the average daily maximum

1-hour ozone concentration over the two-week period;, other air pollutants,

and  additional  explanatory  variables.  •  ~n  these   and  all   other  results
                                                               •
presented  below  air  pollution  was  characterized by  the  readings at  the

monitor nearest each  individual's  house  for wnich data were  available.   No

individuals were included in any of the statistical analyses  in this report

if the nearest monitor  with valid data  was greater than  twenty miles from

their home.

     In equation  (1), the  only air pollutant  included  besides  ozone  was a

measure  of  sulfate  concentrations  during   the  two-week   period.     We

concentrated most  heavily  on  these  two pollutants  throughout  this report

because ozone  and  sulfatas (particularly highly  acidic sulfate particles)

appear  to be  the  two   pollutants  most  likely  to  be  harmful at  levels

frequently observed  in  the United  States today.    However,  considerable

attention was  also  devoted to  total suspended partioulate  matter, carbon

monoxide, nitrogen dioxide and sulfur dioxide  in the analysis.  Ozone  was

-------
                                                    »y   •-•
                                                    •r.   c
                                                                                              .—s      o  -•">
                                                                                            •  O      I*"1  O
                                                                                                       o    •
                                                                                                       O  ft
                                                                                                                                                                                  O  -*
                               a o
                               Cl C
                                                                                                                                                         o r-
                                                                                                                                                         m o

                         O   »
                         O  —
                         O  w
                                                                                                     m o
                                                                                                     ^j on
                                                                                                                                                                                   0
                                                                                                                                                                                   O
                                                    cvj /-N
                                                    p-4 r*j
                                                    -*• O
            O   »
              t  m
              I   w
                                                                                                       »-»  u^

                                                                                                       O  •-•
            ^  a

            o  o
                                                                                                                                    ~     <-i  O

                                                                                                                                                         O O
                                                                                                                                                         O w

  •    »     o
sf  ^J       I
—4  U*V
O    •
o  —
                                                                                                                                            •O  O
                                                                                                                                            'M  O
                                                                                                                                            o   •
                                                                                                                                            o  o
                                                                                                                                                                                                                y-s     O  -^
                                                                                                                                                                          ^      o  *-•

                                              I     —1  ^
                         ^  IX
                         C1   •
                         o  —
                                                                                                                                            —  nj
                                                                                                                                            O   »

                                                    fNj y^
                                                    o  -r

                                                                                                                                            o  o
                                                                                                                                            O  w

                                                                                                                                                                          »-.       O  -7
t      —     I     •-«
                                                                                                                                            ^-  O
                                                                                                                                            O   »
                                                                                                                                                                                               O    t
                                                                                                                                                                                                 •  o


-------
                                                               4-8
AJ i%    r^ f~t
O  •    O   i
 I —    O «
   w    O w
           i
          I

                                                                                             0  I
                                                                                             0 O
                                                                 0   «
                                                                 0  O
                  « >-
                  a. cj?
                                              O  t
                                              0 *4
                                              O w
                                                                 O   •
                                                                 o  a
                                     — o
                                     O
                                                        .»  o    —
                                         I    O   I
                                        -i    O —
                                        w    O w
                            O  >     i   •
                             I O    f  •*
                                               *» o
                                               o  •
                                               O 0
                                               0 w

o •
0 0
t w
•0 -O
O i
o o
C2 >w
^ S 55 ZZ
• t Oi • <
i^J O O •- *C
^^ f w — •
»• >r x 0"
— M <1 »-l
• • O •
1 •« O 0
w O w
IX -a 5 —
• 1 O •
t-« O »«*
w » w
3 O ^
cr. • *^
o — •
• "M -J

O  t
o *~
S w
                                                                                             O O
                                                                                             o *-/
                                                                                                       f  ^-t     O  O
                                                                                                                            s -o
                                                                                                                             * >T

                                                                 00    -< xT
                                                                                                           >     O
                                                                                                          —     O
                                     ^ ^


                                     O
                                        —    O  o
                                              O  w
                                               :*• ^s    sr

                                               f O    «•

                                               6 o
                                                                 0
                                                                 0
                                                                                             — o
                                                                                             c  •
                                                                                       «     o
                                        ^    O  w
                                                           —    c; o    «
                                                                                       -i     o o
                                              0  O
                                              O  W
                                              o
                                                                 o o    —
00
0.
c
o
o
MP
DM

-------
positively related  to the  number  of MRADs  during the  tvc-week  period,  a




relationship  significant  at  the  10  percent  level.    Surprisingly,  the




sulfate  variable  was negatively and significantly related  to MRADs.   We




have no explanation for this finding.




     Of the other  variables in  equation (1), only  three were significant.




They were  income  (INCOMCON),  the  smoking  dummy,  and  the  dummy  variable




indicating the  presence  of a chronic illness or  impairment.   All  three of




these variables had the expected signs, suggesting that higher income means




fewer days of  acute illness, while  smoking and the presence  of  a chronic




illness imply more acute illness.  The t-ratio on the chronic illness dummy




demonstrates the very strong effect it exerts on the number of MRADs during




a two-week period.   The R  for  the  regression  is  0.08  and the 7-statistic




is 22.=.




     Equations  (2)   -   (^)  replicate  this  regression  but with  different




combinations of  air pollution  variables.   In  equation (2}  sulfates  were




replaced  by  total  suspended   particulates.    The  coefficient  on  osone




remained  positive  (as  it  did  throughout  virtually all of  the  analysis)




although  it   was   no  longer   significant   at  the   10   percent   level.




Particulates,  on the  other hand, were positively related  to  MRADs, at the




10 percent  level.   The  coefficients on the  chronic limitation  and income




variables increased in  significance,  while  that on  the  smoking dummy fell.




Additionally, the coefficients on both FAT and FATSQ (the proxy for general




physical  condition)  increased  in significance.   Their  signs  suggest  a U--




shaped relationship between weight-to-height ratio and acute morbidity.

-------
                                   4-10
     In equation (3), ozone was entered with both sulfate concentrations as




well as  total  suspended particulates.   Ozone again was  positive although




not  significant  at  the  10 percent  level.   Participates  were  positively




associated with  acute  illness while sulfates  had a negative   association




significant  at  the  5  percent  level.   The  coefficient  on  the  smoking




variable  was  significant  at  the 5 percent  level.   Equation  (U) included




five different pollutants.  Of  the  five,  only sulfates were significant at




the  5 percent  level, again, however, with an unexpected sign;   particulate




matter was positively  associated with MRADs at  the  10 percent level.  The




coefficient estimates  on  the  other independent  variables  in  (3)  and (u)




were similar to those in equation (1"!.




     In  equations   ' 1 )  -  (4)  the number  of  individuals  included   in  a




particular re°°ression  varied  from  ^  5°''  to  3  175.   This  is  because when




more air  pollution data were  recuired for a  particular  regression,  fewer




individuals were likely to have  them.   The largest  sample size occurred in




regressions with ozone  and  particulate matter because  of  the  large number




of   particulate  monitors   around  the   country  compared   to  the  other




pollutants.




     Equations (5)  - (3) replicate  (1)-  (4)  with one important difference.




Since the relationship  between  ozone and  acute morbidity may be non-linear




(increments to the average daily maximum  1-hour ozone  concentration may do




more harm at  higher  than  at lower  levels),  we reexamined the  determinants




of MEADs  using the square of the average  daily maximum  1-hour concentration




to   measure   ozone  during  the  two-week  reference   period."     As  (5)




illustrates,  the  coefficient  on  the  square  of  ozone   was  of  the expected

-------
                                   4-11
sign and  was significant  at the  1  percent  level.   Once  again,  income.




smoking  and  chronic   illness  were   the   other  variables  significantly




associated with  acute morbidity as measured  by MRADs.   In equations (6) -




(3) ozone was always significant at the 5 or 10 percent  level.




     In equations (9) and (10),  this basic model was reexamined again, this




time using a logit specification rather than ordinary least squares.    Any




individual having at  least one MRAD  due to any cause  during the two-week




period was given  the  value unity,  while those  with no  MHADs were assigned




the value  zero.    (RADPRESM  connotes  the  presence  of at  least  one MRAD,)




Equation  (9)  relates  in a  linear fashion  the log of  the  odds  that  an




individual had at least  one  day of acute illness to the set of explanatory




variables.     As   the   t-ratio   indicates,   ozone  was   positively  and




significantly  'at  the   5  percent  level)  related   to  the" likelihood  of




experiencing a MRAD.   The regression  suggests that  women are  more likely




than men to experience  at least one MRAD,  as  are smokers, those with lower




incomes,  individuals  with a  chronic   illness,  and  those  living in  areas




where precipitation is relatively high.  The measure of  persons per  room in




the household was also significant  at  the 5 percent  level although the sign




is not  as  expected.   Equation  (10) replicates  (9)  using  the square of  the




average daily maximum 1-hour ozone concentration.   As in regressions (5) -




',3K this change increased the significance of the  coefficient on ozone,  in




this  case nearly  to  the 1  percent  level.    In   both (9)  and (10)  the




coefficient on sulfates, while still negative, was  no longer significant.




     Table ^-2  presents the analysis  of  work  loss and bed  disability  in




adults due to all causes.  Equations (11) - (13) experiment with different

-------
4-12
Table it- 2.
Model #
Dep. Var

Ir.dep. Var.
INTERCEPT
03 JIB 01
S4NR01
SFNR01
N2NR01
CON?. 01
a ACCT 130
SSXM1FO
MARY 1 NO
INCOMCON
CITYI100
FAT
FATSQ
AGE
Work Loss and Bed Disability Among Adults:
(11)
*
WKSLLD2W

.403
(0.73)
-.691
(o'.65)
-.0014
(0.33)
-
—
—
.0^62
(0.65)
-.0988
(1.88)
.0*111
(0.77)
-7.336E-7
(0.36)
.0172
(0.37)
-.273
(0.72)
.0632
(0.87)
.01"U
(1.39)
(12)

WKSLLD2W

.521
(0.94)
-.1U8
(0.42)
-.000245
(0.06)
-.00114
(1.41)
—
—
.0^5"
(0.64)
-.0976
(1.35)
.0^18
(0.78)
-9.952E-7
(0.35)
.0197
(0.42)
-.287
(0.76)
.0659
(0.91)
.0146
(1.41)
(13)

WKSLLD2W

.562
(0.86)
-.671
(0.55)
-.00222
(0.40)
-.000706
(0.74)
-.OC03<2
(0.56)
.00760
•(0.73)
.0758
(0.92)
-.'•53
(2.55)
.0501
(O.SO)
-4.323E-7 -
(0.13)
.000277
(0.01)
-.251
(.057)
.0731
(0.86)
.00614
(0.5D
(14)

3EDDIS2W

1 .065
(2.13)
-.498
(0.49)
.00238
(0.60)
—
—
—
.115
(1 .68)
-.115
(2.40)
-.0169
(0.33)
.00000612 -
(2.33)
.0563
(1.25)
-.813
(2.36)
.203
(3.06)
-.00732
(1 .07)
All Causes
(15)

3EDDIS2W

1.173
(2.33)
-.248
(0.2a)
.00361
(0.38)
-.00118
(1.59)
_,_
--
.115
(1.68)
-.11-
(2.37)
-.0162
(0.32)
.00000637 -
(2.42)
.0536
(1.30)
- .323
(2.38)
.205
(3.08)
-.00759
(1.04)

(16)

3EDDIS2W

1.344
(3.21)
-.419
(0.37)
-.000301
(0.06)
-.00138
(1.62)
.000652
( 1 . 1 9 )
.0015*4
(1.7^)
.129
(1.53)
-.0852
(1.56)
-.0338
(0.59)
.00000357
(1.20)
.106
(2.04)
-1 .510
(3.87)
.345
(4.57)
-.00797
(0.97)

-------
4-13
Table 4-2 (Continued).
Model *
Dep. Var.
Indep . Var .
AGESQ
SMOKY 1 NO
EDCCMCON
CHRLMDUM
*
CROWDING
AVMAXTM?
AVPRECI?
HUMID RF
INDOCCAV
R2
? or X2
N
(11)
WXSLLD2W

-.000221
(1.82)
-.0164
(0,35)
-.0204
(2.2U)
.396
(4-. 78)
-.0^38
(0.9*0
. C 0 " 0 4
(0.63)
-.214
(0.93)
.00461
(1.63)
-.0202
(2.38)
.0186
2.83
2,854
Work Loss
(12)
WKSLLD2W

-.000223
(1.32)
-.0149
(0.32)
-.0205
(2.25)
.398
(4.80)
-.0658
(0.84)
.00113
(0.68)
-.274
(1 .17)
.00381
(1.3D
-.0207
(2.41)
.0193
2.78
2,351
and Bed Disability Among Adults:
(13)
WKSLLD2W

-.000127
(0.89)
-.0141
(0.23)
-.0250
(2.35)
.335
(3.47)
*
-.133
(1.51)
.00272
(1.38)
-.290
(1.04)
.00397
(1.12)
-.0185
(1.33)
.0210
2.04
2,107
(14)
3SDDIS2W

.0000758
(0.99)
.0614
(1.36)
-.00985
(1.28)
.7^6
(11 .7D
.00219
(0.03)
^ P 1 0 Q
— . >j o , ^ V
(1.25)
-.304
(1.36)
.00561
(2.05)
—
.0483
14.50
5,200
(15)
3EDDIS2W

.0000737
(0.96)
.0625
(1.37)
-.0102
(1 .32)
.719
'11 ^ 4 ^
.00693
(0.09)
-.00191
(1.20)
-.373
(1.64)
.00488
(1.75)
—
.0488
13.96
5,193
All Causes
(16)
3EDDI32W

.0000934
(1 .08)
.0272
(0.52)
-.0150
(1.74)
.733
(10.59)
-.00292
(0.04)
-.OOOQ30
(0.5D
-.0505
(0.19)
.00381
(1.15)
—
.0589
11 .61
3,918

-------
                                   U-1U
combinations  of  air  pollutants   for   adult   work  less,  while  similar




combinations are used in (1U) - (16) to explain adult bed disability during




the  two-week  reference  period.   As indicated,  none of  the  air pollution




variables was significant at the 5 percent level in any of the regressions,




and only carbon monoxide (CONR01) was significant at the  10 percent level—




in  the  regression  explaining  adult bed disability.   The H'~'s  were much




lower in  these equations  than those  explaining  MRADs.   Substituting the




square of  ozone made no difference in any of  these results.   Of the other




explanator3/  variables,   the  chronic  illness   dummy was still  the  most




important.  It was highly significant in all the equations,  Equations (11)




-  (13) suggest that women experience more work loss and bed disability days




than men,  and  tha~  the  amount  of work  loss  is negatively and significantly




related to education.




     Of particular  interest  in  the  work  loss  regressions is  the measure of




paid  sick  leave,  INDOCCAV.  It  was significant at  the  5 percent level in




two  of the equations and at the  10 percent  level  in "he  third.   Its sign




suggests that  work loss varies  inversely with paid sick leave.   This may




reflect  the  willingness  of  workers  with ample paid sick leave  to  miss an




occasional  day of  work in  the early stages  of illness, thus  preventing




prolonged  absences  due  to  illnesses that are  allowed to grow more serious.




It  may also  reflect the fact  that  the least  physically demanding jobs (in




offices,  for  example)  often  carry with  them generous  paid sick  leave.




Thus, the  measure of paid sick leave may actually be distinguishing between




relatively safe  and "clean" jobs  and  those where  hazards may be present.




To  explore this  possibility,  an  additional  work loss  model  was estimated

-------
(not presented here) in which  a  blue  collar dummy variable was substituted




for INDOCCAV.  The  dummy variable was positively  related to work loss and




was significant at  the  1  percent level, thus  giving support to the latter




suspicion.




     Because   of   the   positive  and   often   statistically  significant




association between ozone and  MHADs due  to  all causes in adults, a variety




of sensitivity analyses  were  performed to  determine  the  robustness  of the




relationship.   These  are  presented  in Table  U_3.    In  equation   (17),




attention  is  concentrated  on  a  subset  of  individuals  for whom  data are




available  on  the  usual  set of  regressor.  ,  but  also for ragweed (pollen)




concentrations."  Equation  (17)  is  identical  to equation (1 )  save for the




pollen  variable.    The  introduction  of  this  variable   (i)  reduced  the




magnitude  of  the  coefficient  of  the ozone variable;    (ii) reduced the




t-ratio on  the  ozone  coefficient from *. 92  to '.12;  and  (iii) reduced the




sample size by more than half because pollen data were only available for a




subset of the full sample.




     In  (18)  - (20) ozone  was  measured not by the  average  daily maximum




1-hour  concentration  over  the two-week  period  but  rather by  the average




hourly  concentration over that  same  period.  (Rather than  average  the *, H




daily highest one-hour readings,  all hourly  readings up to a maximum of 336




were averaged.)  This change reduced the significance of  the ozone variable




when  compared  to  equations  (1)  - (U).   This is  not  surprising, however,




since   it   is  commonly   thought that   peak  rather  than  average  ozone




concentrations are most likely to induce or  exacerbate acute morbidity.

-------
                               4-16
Z

s:
ee

z

£
                                                      in ^s
                                                      O 0^
                                                               r~\    *C *•>
                                                                                         — o
                                                                                         o  u
                                                                                         o o
                  O ^
                  O 0*

O


o
              O    — O
              O   -^ —

-------
                                              -f-i/
                  o  <*^
                  X*  O
                  r-j  >y
                  o   »
                             o  ^     o ^
                             O
                             O
                             O
                             O
                             O
                      rg  -O
                       *   •
                       I   O
                                 m r*
                                 \f- so
                   cr  ff-
                   o   *
                   o  o
o   .
o  o
O  w
X
Q
                                                                        e*  a
                                                                        o   •
                                                                        C1  "-•
                                                      03  C=
                                                      C   .
                                                      o  o
                                                                                                                     * •-•     —•
                  I-NJ  -J-
                  o   •
o  *-*
tf\  flO

o   *
O  j j*^
:  lf\     U^ P^
  O     tA -fi
   •     o   •
  O      • O     00
                                                                                   s,T  %O     r*
                                 *f /"^
                                 14-1 CD


                                 O   •
                                 O Q
                                                                            W     O  ^^
                  C>  O     1-1
O   i      "   •      •   •
O  fvj      1   —•      I  —<
a  -w         '«-'        v>
o
 *
 !
                                                                                   f- ^     O <-*
                                                                                   LTi v?     O.1 W^
                                                                                   — o     -H a
                                                                                                                   o  o     «—
                             cr  /^
                             *-»  >*•
                             IM  1f\
           o  o
           ^  ^.
           c   •
            •  o
            I   ^
                                                                                   o   »
                                                                                   o o
                                                                                   O W

                                                                           -^     o —
                                                                                     .CO      xf  C
                                                                        CJ  C?      CO
                                                                                                                   o    »
                                                                                                                   o  o

                             o
                             o
                             o
                                                  i^-v  ^,     u i  i*j     T' O

                                                   1O     O  C     O —•
                                                                                   o o
                  i-»  "SJ      -4
                     \^      O w
                                                                                  O  w
                                                                                  o

-------
--13
I 1
1 !
t i
! ',
1 )
1 1 /^
1 1 r-
1 1 iSJ
1 1 w
1 !
1 1
i i
; |
l t
I !
1 )
1 i ^
1 i o
! 1 rg
1 1 w
1 t
1 l
1 <
1 |
1 1
1 t
1 1
1 1 >**
I 1 lA
1 1 "J
| | ^s
> I
\ \
t 1
1 1
I 1
! 1
1 ! *>
1 1 s*
i irt i "M
! -,• I x^
1 in i
1 3 !
f *T> t
1 LJ t
j j
1 — * <
t — 1
t «r 1 ^
1 ! -M
> .. i w
1 ^ (
i *" !
' — « 1
i 3 l
l < t
I C l *>
1 — 1 -w
1 1 , "%
1 > ' W
, f i
— !
i > 1
! »" t
1 o '
| «* ; ^
7 ! ""
1 ** ! w
i •_• l
i "* i
1 U 1
•- I
1 i- i
' T 1
1 1 '**
» *- 1 C?
1 O ! 'w
1 1 -^
I 5.1 1
! t 1
! yi 1
l >, i
1 — 1
t *" i
i C t
«— f >™s
i i r-
l > ! —
1 *- ! •M'
1 •- 1
1 > 1
! -* 1
' *" !
i U' !
I C I
I li f -^
] . « | ~ >
1 1 —
I 1 W
i » 1
1 -"t I
1 T* i
1 t I
1 3 1
! C !
1 -< 1 •— i
! *-• 1 ^
t C 1 **
1 0 1 w
1 U 1
' w 1
f 1
t — 1
1 i 1
1 sj" 1 f
! I
t t) i — «
1-41*
! r i —
1 *3 1 O
i «~ i r
t



C:
2T
«— i
s:
c.
r*
a.



a:
c
2
k— 1
s.
"
<
a:



CK'
a
*!
i««
J"
a
«i
Ci.


a:
D
^
>-•
Z
cs
ft
zs



&•
o
z*
r
.~v
 a •
^ u
— i -t
c >
*-»
\
i
I O /^ «* /^i (T> >-*
1 in o CT- ^ -» cr-
1 C l * • O *
100 — o —
! t w w t w
I t
1
t
!
I
l ar /-* — ^ o /•>
j jr- »*» ^ pi- Lf^ r^
1 -O *^ VT* O1 *N. t*-
! 0 • » . 0 «
1 O *M *-« O O
1 * N^ S> • ^>
1 i 1
1
I
1
t
1 >£• (""N >O ^^ ^ ^^
( «C CO 0, "J *» f^
! ~i mi vT (^ rn o
10* t • 0 •
100 -« O -*
( O s-» v-* * ^/
> • t
1
f
1
1
! /*i -r *o ^ -v; •*-
1 ••* O ^^ <•"" ("^ kT»
1 0 • . . 0 »
too o o o
low vv • ^
1 * 1
I
t
1
J
1 O *** IT1 '^ — • ">
j »y rn ^ -C ^* ^*
1 «4 l/> P«- O -H -",
; o • • • o «
< 0 O ry c* O
3 )
1
1
I
1
I CD ^ CC '^ C^ ^^
1 ^ ?! ^ ^ I ^
! 0 C* 10 00
• w w O w
i ! •
1
1
!
I
: * s* 0 ~ f • ->
; ^- L" ^ T-> PVJ ^j
O • i 0 .
i f
1
I
i

! *"* T, ^ ^ ^ ^
1 ^ *J f* "**! W"l •*
' O • • • O t
1 O O O O —
i C? w w • i-/
! * i
i
1
1
1
! -f" *** -f '•"v ••— ••"'*
1 •-* O IT « -^ i-*
i ^-« -r -^ v: w -*
1 0 « . • 0 t
I O O — * O O
i 1 ^ W » W
1 1
1
1
t
1
1 rvj ^ ^- ^ ."v ^
1 — r— •») r\* rr-, o
10. i • 0 •
100 -• 0 •*
1 • 'W V^ « v-i
1 f
1
!
!
|
1 o --\ ,-i ^» »•— ^^
| r- "1 fj Oj ^ MA
1 "" O r«j >j •-( ("J
! O • « • O •
! o c> o on
t o s-; w • ^-«
I 0 1
i »
!
1
t
1 C. u*
1 v" ^^ tL
1 ^ o ac
! ^ i!^ r^
i «a ^ •—
15" ft ET
i > > 3



r« ir* »-«
0 ^C ^
00-







y-\ O r\j
^ ^- f"1
^- * * t rj
O •- •
* ~* r**





03 (?* rvj
O « U'
"^ -^ u^






"" ° "~
T . M
o ^ -





u y ^
X "i lA
C-3 tJ* *
• — -•






r^ ir\ M
0^ sj- w
^*. * "M
O "•* *
• •*• ' -J-






*r* 3 Q;
L "^ J2
O rvj •
• r\j <*•






— iuT ^»
^r fn fc-
r«v f r*j
O fN »
* — . r-j







"'•I
X

i_
o
-^
Cjf U. 2

-------
     Equations  (21)  -  (2'J)  replicate  equations  (1)  -  (U)  with  one




difference.   The  sample of  individuals  in (21) -  (2U" were what  we term




"self-respondents."   That  is, during  the HIS  interview  these  individuals




were present and answered the interviewer's questions for themselves (if an




adult was not present during the  HIS, his or her spouse—or other available




adult—provided  the  relevant   information   for   the   absentee).     Self-




respondents  were  singled  out on  the presumption  that  information  about




their acute  morbidity might  be more accurate than  if  another  had provided




information  for  or  about   them.     Across   the   four  regressions,  the




coefficients  on  the  ozone  variable  were  generally  smaller  and  less




significant  than  in the larger sample that included both self-respondents




and  those  for whom  information  was provided by  others.   This  casts  some




doubt on  the earlier positive and significant associations between  ozone




and  MRADs.   We cannot  rule  out  the possibility thac  the  smaller  and less




significant  coefficients   are  due  to  the  smaller  sample size,  however.




Other sensitivity analyses presented later  in this  chapter make use of the




distinction between self-respondents and others.




     Possible  interactive   or synergistic effects of  ozone  with  other




pollutants  or meteorological  variables  were  also  explored.    In  equation




(25),  both  the  average daily maximum  1-hour  ozone  concentration and  a




multiplicatively interactive term involving ozone and sulfates  were used as




explanatory  variables  along with  particulates  and the other variables  in




the  basic  model.     3y itself,  ozone  was  positively  and  significantly




related to  MRADs.  However, the  sign  of  the  interactive  term was  negative




and  significant.   In equation (25) ozone was allowed  to  interact  with the

-------
                                   U-20
average daily maximum  temperature.  Although  the coefficient  on ozone in


this  model  was  negative  and  significant,  the  interactive  term entered


positively and significantly.


     With the exception of the logit regressions and those using the square


of  the  ozone  concentration,  the analysis  to this point  has  been based on


linear regression over  the whole range of ozone  concentrations.   Yet many


clinical  studies  are  designed  to  determine  whether  some  "threshold"  or


lowest  observed  effect  ozone   concentration  exists,   below  which  health


effects are not detected.   Accordingly, several regressions were run using


"switching"   techniques  which  permit   different  relationships   to  be

                                           7
identified for  different ranges of  ozone.    Equation  (27),  which had the


highest F-statistic of the switching regressions we tried, presents results

              k
from  a regression  where different  relationships  between ozone  and acute


morbidity   could    be   identified   for  average   daily  maximum  1-hour


concentrations  above  (03H15)  and  below   (03L15)  0.15  parts-per-million


(ppm).


     In (27), for average daily maximum 1-hour ozone concentrations greater


than  0.15  ppm,  the coefficient  on the ozone  variable  was  3-62 (the slope


measure) and  it was significant at  the 5  percent level.   Below 0.15 ppm,


however,  the  slope  was  much flatter  (0.3)  and  the  coefficient  was not


significantly  different  from  zero.    This  suggests  that   a  different


relationship  may  exist between  ozone  and  MRADs from  all  causes above and


below  that level,  a  finding consistent  with clinical  research relating

                                7a
ozone exposure to lung function.    Regressions were run using  different


switch  points both  greater and less  than  0.15  ppm but in no case was there

-------
                                   U-21
such a clear difference in slope and significance between the two regimes.


     Summarizing  this  analysis  of  acute  morbidity in  adults  due  to all


causes, several  points  emerge.   First, regardless  of  how it was measured,


the  coefficient  on  the ozone  variable was  positive  in  every regression


explaining minor restricted activity days in adults.  It was significant at


at  least  the  10 percent  level  in about  half  the  models, although its


significance did fall  when the  sample  was  restricted  to self-respondents,


for  whom  reported  morbidity  data  were  perhaps  the  most  accurate.   No


evidence  was  found linking  ozone to work  loss or bed  disability.   Total


suspended   particulates  were   also   always   positively   and  sometimes

                                                          q
significantly related  to minor  restrictions  in activity.    Sulfates were


consistently  negative  and   significantly   associated  with   MHADs,   a


counter-intuitive  finding.    Of  all  the   regressors,  the  dummy variable


indicating a chronic illness was by far the most significant.




     4.1.2  Children


     Children are of  particular importance because they are thought to be


more  sensitive  than   adults  to  the  effects  of  ozone   and  other  air


pollutants.^    This  is  not  only   because  their  bodies   are  developing


structurally, functionally,  and biochemically, but also because they tend


to be outdoors more than the general population, exercise more heavily, and


ventilate more   per  unit body weight  than adults.   Using  subsets  of the


15,711  children  in the  final  RFF data   base,  a number  of  models were


explored  linking varying  types of  acute   morbidity  to  air  pollution and


other determinants.

-------
                                   4-22
     The  results  of  the  analysis  of  children's MRADs  due to  all causes




during  the  reference period  are presented  in Table  4-4.   The  number of




explanatory variables used  in the  analysis  is smaller  because  height and




weight  were not  recorded  for  children  in  the HIS  (thus  necessitating




exclusion of  FAT and FATSQ),  nor  is  the  marital status  or  smoking  dummy




relevant  (because  no one  under   the  age  of  17  was  given the  smoking




supplement).




     Equation (28) includes ozone and  sulfates as the relevant measures of




ambient  air  pollution.     Although  both  had  a negative  and  therefore




unexpected  sign,  neither  was  significant.   Boys  had more MRADs during the




reference  period  than girls,  a  finding  significant  at well above  the  5



percent level.  The  coefficients  on AGE and  AGESQ suggest that MHADs  among




children  decrease with age  from  birth to' about ten years and then  increase




between age ten and age sixteen, the point at which youths become adults in




the HIS  framework.   Once again  the chronic limitation  dummy  was  the most




significant determinant of acute morbidity as measured by MRADs.




     Equations  (29)  and  (30)  reproduce this  basic  model  using different




combinations of air  pollutants.   None  of  the  pollutants was significantly




related  to MRADs  in children.    Equations   (3D  -  (33) reiterate  this



approach using the square of the average daily maximum 1-hour concentration




as the measure of ozone.  Unlike the adult regressions, use of the squared




term had  little effect.   Over  the six different models  estimated in (28)  -




(33),  only  nitrogen  dioxide  showed a  positive  association with  MRADs in




children, a finding not significant at  the 10 percent level.

-------
4-23
Table 4-4.
Model #
Dep. Var,
Indep . Var .
INTERCEPT
03NR01
03NR01SQ
S4NR01
SPNR01
N2NR01
CONR01
RACEW1BO
SSXM1FO
INCOMCON
CITYI100
AGE
AGSSQ
CHRLMDUM
Restricted
(28)
RADMINOR

.165
(1.24)
-.802
(1.33)
—
-.00284
(1.29)
—
—
—
.0178
(0.51)
.0633
(2.51)
4.444E-7
(0.29)
"-.0161
(0.58)
-.0476
(4.74)
.00225
(3.83)
.375
(5.41)
Activity
(29)
Among Children
(30)
RADMINOR RADMINOR

.160
(1.16)
-.819
(1.35)
—
-.00287
(1.28)
.0000604
(0.14)
—
—
.0178
(0.52)
.0628
(2.49)
4.438E-7
(0.29)
-.0159
(0.57)
-.0477
(4.74)
.00226
(3.83)
.375
(5.41)

.184
(1.19)
-.684
(1.03)
—
-.00289
(1.08)
-.000110
(0.23)
.000446
(1.41)
-.00329
(0.64)
.0107
(0.23)
.0497
(1.76)
.00000183
(1.06)
-.0173
(0.54)
-.0496
(4.39)
,00246
(3.71)
.369
(4.71)
: All Causes
(3D
RADMINOR

.171
(1.28)
—
-2.692
(0.84)
-.00306
(1.40)
—
—
—
.0167
(0.48)
.0633
(2.51)
4.812E-7
(0.32)
-.0140
(0.50)
-.0476
(4.74)
.00225
(3.82)
.376
(5.42)
(32)
RADMINOR

.168
(1.22)
—
-2.736
(0.84)
-.00306
(1.37)
.0000345
(0.08)
—
—
.0167
(0.48)
.0629
(2.49)
4.718E-7
(0.3D
-.0137
(0.49)
-.0476
(4.74)
.00226
(3.83)
.376
(5.42)
(33)
RADMINOR

.195
(1.25)
—
-1.849
(0.55)
-.00306
(1.14)
-.000141
(0.29)
.000397
(1.27)
-.00303
(0.59)
.00952
(0.25)
.0496
(1.75)
.00000186
(1.08)
-.0146
(0.46)
-.0496
(4.39)
.00246
(3.71)
.370
(4.72)

-------
4-24
Table 4-4 (Continued).
Model #
Dep . Var .
Indep. Var.
CROWDING
AVMAXTMP
AVPRECIP
HUMIDRF
R2
F OR X2
N
(28)
RADMINOR

-.0253
(0.65)
.00105
(1.14)
.111
(0.84)
.00149
(0.94)
.0131
5.58
5,480 '
Restricted
(29)
RADMINOR

-.0257
(0.66)
.00106
(1.14)
.114
(0.86)
.00150
(0.94)
.0131
5.16
5,473
Activity Among Children: All Causes
(30)
RADMINOR

.0223
(0.53)
.000279
(0.26)
.223
(1.46)
.000193
(0.10)
.0133
3.44
4,096
(3D
RADMINOR

-.0262
(0.67)
.000573
(0.71)
.127
(0.97)
.00143
(0.90)
.0129
5.'50
5,480
(32)
RADMINOR

-.0265
(0.68)
.000575
(0.71)
.129
(0.97)
.00143
(0.89)
.0129
5.08
5,473
(33)
RADMINOR

.0218
(0.52)
-.000176
(0.19)
.236
(1.55)
.000177
(0.10)
.0131
3.39
4,096

-------
                                   4-25
     Table 4-5 illustrates the results of the basic model applied to school




loss among children.  While the coefficient on ozone was positive equations




(34)  -   (39)  its level  of  significance  was  always  very  low.   With the




exception  of  nitrogen  dioxide,   none  of   the  air  pollutants  played  a




significant  role  in explaining children's  school  loss.  In  both equation




(36) and  (39)  nitrogen dioxide is positively  associated  with school loss,




an effect significant at the 5 percent level.




     Among the other variables, race  was  generally significant at at least




the  10  percent  level;  the  dummy variable  indicates  that white children




experience more  school loss  due  to  acute  morbidity than  black children,




holding  income  and  other  factors constant.   As  usual, the  presence  of  a




chronic illness has the  strongest association  with acute morbidity leading



to  school   loss.       Average  maximum   temperature  was  negatively  and




significantly related to school loss.  This may reflect a reduced incidence




of influenza and other  flu-like diseases  in warmer areas during the school




year.  The dummy variable for central city residence was significant at the




10 percent level  in almost  all  these regressions,  indicating more school




loss days for central city students than for those outside the central city



but still in an SMSA.



     In equations (40) and (41) logistic regression was employed to examine




the sensitivity of  the  school  loss  results  to  the  method of ordinary least



squares.  Neither ozone  nor  other air pollutants  played a significant role




in explaining the variation  in school loss  days.   As in the ordinary least




squares estimates, the race of the student and the presence of a chronic

-------
4-26



























u
"*
3
LJ
,-*
<

"
&
L.
T)
•— t
U>
r*
C
O
E
*T
i/»
Ift
O
J

*w
O
0
r
0
i/j


•
in
i
^f

9

A
nj


**s
«M4
>*
w




/«»
o
<*
W



CT>
m
W



/^
m
m
w



r^
r—
m
W




r>
•O
fl





*>
m

w





/•>
sj
f)
W









^».

^
*

O
a:
1/1 /-s
Ui *~
QC »-»
O. • 13
-1 Q
OO -J
2 s^


UJ N-
QC *-*
a o
_j a
*/i _j
3 w
3
rg
j

2

3E
O
-J
J
VI
it

3
Q
_J
_j
f>
:*
2

2
a
j
j
L/"»
^
2


3
t\j
C
J
-J
*/)
^
3

3
r j
0
-J
_J
u^
X
t


•
u •
fl) k
> 11
>
•
a. •
Or Q.
o *
TJ
C
>-H
O f*» r*» ^> ( »O ^
^5 tn •* O 1 ^ f*j
• r»- m »^ f^> ^
rvj • * < O •
1 rg ( o O O
\^ w * w



m f* r- o t »# **
F*. ** ^ .« (N| ftj
« • • • Of
*M (Si 1 O Q O
| W W » \^

en ^» 1 %O r* p< r*
• * • • 0 •
<- m o o o
o
t
t
P\/ P* | CP Ol ""\j m
vr **• cr« •-*• ^ HI
f • . » O •
^ U^ O O O
\S s-/ • %-•


_  u*\ ^* -*'%<•.
• • t • ei »
-^ in o o O
^^ N-' • \-/
i


m -* o *3 ( o f^
r^. or <•*» u". «• O
• • • • O •
— o Q a
.
i


in en o ft | <^ •*
in •* o r-  ^< | f. ^,
%r *O f\i in i *N •*
in m O r^ f\j g^
• • • * 0 •
~i — O O £3
^^ v/ t s*/
t





H-
0. O
IU t>l
U n r-. ~*
Of O O O
UJ QC QC C£
i- ? Z ^
2 (*i tn ^
i-< a a to
*r ^ l
»4 »*• t
o* m
0 •
0 0
O v->
•
1

1 1





0 * O •
O O O rst
• W
t

•^ ^ 1
O •
o o
O w


t |






i-> O -« O
O • O t
a o o <^i
»
1


•j?  ^s i
t fM N. —i PH. nn m |
>O 1^4 rg r* CNJ m
t • • t O •
m if* o o
w w O ^
O
O
1
t CNJ a* «w s*- 1*4 m i
«O  f«-*^ m»* ^.rt.
-*i^ x»«^ «-» */" a:c:.'
• t Of O • 0 •
— • -« 00 • -*
w 1 w O w w
o
1
! — -J- rrWj '/*u- f^-O
«£* «J *^ -^x. CD-^T
• • r? * T» • "» «
~ • -^ o o » -^
^^ 1 vs C* W ^>
O
,
1
f*- (^ ^-O «O"^ ^1*^ »>fcf*^
Ot O • Ot O* O *
ora *— i^« oci < —
1 0
o

1
1 -*c* P—m u"> — * i^r-j
«CT» ^r*-« «"••*; o^r-
• • o • o» c«
^ • ^ o a • «-«
s^ i |w r^ ^-^ v«/
o
o
1

i in^s ^.^s O'*** o^>
1 «H rg an O -«C •* C* f"i
«*rt* >#»- —» \o arc^
• • o* o* o*
-4 t *^ O O t -«
s^ 1 s-^ O W ^^
O
o

I



o z o
iTJ O IT n
fN •-* t^ (_?*•«
0 S •"• X >-•
at uj z a >-
3 W X t_> h-
a «t uj z •-»
U O£ L/l fc^ LJ
CO ^ O ^ f*J ^
IW *•» f» f> O^
ao u i mm ^^ rg
o * f> • » •
• 0 O O -H IA
S^ f S-* V-*
1


«S4 ^- O ^ m TJ
N- ^ m ^ ^- -J
o • o • • •
• O CO -. if
v » \^ *-•
1

o •  (•*
O O 
1 >— ' O ^ N-X


^ ) ^^ — v"1 r- *•
c> » r^ • • •
i O 0 O *-
1 ^^ "^ ^-/ \^
•


^ - - £ ~ -
o • o • • *

t w w **x



O ^* •-* ^ r - "
r^ !~t •« n fk. <-j
O • C3 • • •
• O 0 0 P-
1 v^ O w^ '^^
•


?-,?*> O "^ »— i*»
rsj fl C^ w f C.
«M rg t.~i r\j r- ""
O * C- • • *
* 0 C 0 »*-
1 W O W '*-'
•






z
3
o
<^ E
^ -J
U_ lU QC
o o r
•^ «* LJ

-------
                                                            4-27
o
o
.c
                  a.  L;
                  -i  a
                      3


                      O
                         •o
                         c
                                              o   •
                                              o  —
                  O   i
                  o  o
           i ^>    a>  <^

           'CO    O  »*
              .    o   •
            o    o  o
                                                                                ** 00
                                                                                  I 0!

                                                                                O  •
                                     0  •    O
                                      • •«    o
                  o   •
                  o  o
                                              at    o   •    o  •
                                              O  «J    O  O    O O
                                               •  w    O  w    O w
                                               I          t         •
                                                                             **>*   • ff

                                                                             O -*  •
                                                        1
                                                                  1
                                     (J* i*>    f^  ^»    UN  ^*S    ff* ^N



                                     T««    C»»    O*    Oi
                                      to    o^vt    oo    oo


                                               I         I          •
                                                                  I
                                     £**•*    C5X^     ^J^%    O/^
                               rs«  • £X)
                               O V*  •
                                                                             in — or
                                                                             IA -^ ff*
O  »    O   t    O   t

O fM    O  O    O  O
                                              o

                                               I
                                              a.
                                              j:
         o   •    o   •
         o o    o  o
          i ^    O  ^
                                     a
                                     2
                                     a
                                                                 K

                                                                 O
                                   L.
                                   O


                               x u. :

-------
                                   4-28
limitation  were  the  most  significant   factors,   with  average  maximum




temperature significant at the  10 percent level.




     Concluding the  analysis of  acute morbidity  in  children  due  to  all




causes, Table  4-6  presents the results  of the analysis  of  children's  bed




disability  during  the reference  period.   Equations  (42)  -  (44)  utilize




several combinations of  other  pollutants entered with ozone  in its linear




form.  In (45) - (47), the square of ozone is utilized.  Although ozone was




positive in all six regressions,  it was never significant.  Nor were any of




the  other  air  pollution variables, including  nitrogen  dioxide  which  was




found to be significantly  associated with  school loss  days—a less serious




form of acute morbidity  than a bed  disability day.   In the analysis of bed




disability, age and  its  square were both  significant, as  was the presenqe
                                                                •



of a chronic limitation.   Average maximum  temperature  during the reference




period  was  also   negative  and  significant,  as  in   the   school  loss




regressions.  This  again may be due to  the  relatively lower likelihood of




influenza in warmer  areas.  Finally, the  central city residence dummy was



again a significant  determinant  of  bed disability,  almost always at the 5




percent level.








4.2  Acute Morbidity Due to_ Respiratory Disease




     Respiratory disease is  that  type  most likely to  result  from exposure




to  ozone  and  other  air  pollutants  in  ambient   environments.    Before



undertaking  the analysis  of  acute respiratory  morbidity  in  adults  and




children, an additional  requirement was  imposed on  the air pollution data.



Besides requiring  that  any  ozone  or other  air pollution data, be derived

-------
4-29
Table 4-6.
Model #
Dep. Var.
Indep. Var.
INTERCEPT
03NR01
03NR01SQ
S4NR01
SPNR01
N2NR01
CONR01
RACEW1BO
SEXM1FO
INCOMCON
CITYI100
AGE
AGESQ
CHRLMDUM
Bed Disability Among Children: All Causes
(42)

BEDDIS2W
.826
(6.42)
.874
(1.49)
—
-.00214
(1.02)

—
—
.0481
(1.45)
-.000335
(0.01)
-2.024E-7
(0.14)
.0550
(2.05)
-.0304
(3.14)
.00147
(2.59)
.486
(7.81)
(43)

BEDDIS2W
.796
(5.97)
.848
(1.43)
—
-.00246
(1.14)
.000261
(.062
—
—
.0476
(1.41)
.000910
(0.04)
-2.756E-7
(0.19)
.0527
(1.96)
-.0307
(3.17)
.00150
(2.64)
.489
(7.84)
(44)

BEDD9S2W
.873
(5.39)
.931
(1.33)
—
-.00343
(1.22)
.0000150
(0.03)
.000450
(1.36)
-.00273
(0.51)
.0613
(1.52)
.0133
(0.45)
6.736E-7
(0.04)
.0754
(2.26)
-.0430
(3.65)
.00215
(3.11)
.603
(7.86)
(45)

BEDDIS2W
.820
(6.36)
—
3.223
(1.03)
-.00192
(0.92)
—
—
—
.0493
(1.46)
-.000387
(0.02)
-2.470E-7
(0.17)
.0528
(1.97)
-.0304
(3.14)
.00148
(2.60)
.485
(7.79
(46)

BEDDIS2W
.787
(5.88)
—
2.913
(0.92)
-.00226
(1.05)
.00286
(0.68)
—
—
.0487
(1.45)
.000840
(0.03)
-3.113E-7
(0.21)
.0503
(1.88)
-.0308
(3.17)
.00150
(2.64)
.488
(7.83)
(47)

BEDDIS2W
.858
(5.29)
—
2.604
(0.74)
-.00320
(1.14)
.000555
(0.11)
.000513
(1.56)
-.00310
(.058)
.0627
(1.56)
.0135
(0.46)
2.089E-8
(0.01)
.0716
(2.15)
-.0430
(3.65)
.00215
(3.10)
.601
(7.85)

-------
4-30
Table 4-6 (Continued).
Model #
Dep. Var.
Indep. Var.
CROWDING
AVMAXTMP
AVPRECIP
HUMIDRF
R2
F or X2
N
(42)

BEDDIS2W
-.0142
(0.37)
-.00513
(5.74)
.233
(1..85)
-.00506
(3.3D
.0197
9.30
6.018
Bed Disability Among Children: All Causes
(43)

BEDDIS2W
-.0167
(0.44)
-.00520
(5.81)
.246
(1.92)
-.00466
(3.00
.0199
8.70
6.008
(44)

BEDD9S2W
-.0423
(0.94)
-.00592
(5.39)
.313
(1.97)
-.00495
(2.54)
.0271
7.81
4.495
(45)

BEDDIS2W
-.0131
(0.34)
-.00463
(5.99)
.219
(1.74)
-.00500
(3.26)
.0196
9.21
6.018
(46)

BEDDIS2W
-.0157
(0.41)
-.00471
(6.06)
.232
(1.81)
-.00457
(2.94)
.0197
8.61
6.008
(47)

BEDDIS2W
-.0414
(0.92)
-.00531
(5.58)
.296
(1.87)
-.00491
(2.51)
.0269
7.73
4.495

-------
                                   4-31
from monitors  no further  than  twenty miles  from an  individual's  home, a




requirement was added concerning the  completeness  of the ozone data during




the two-week reference period.  Approximately 400 individuals were excluded




from the following  analyses  because they were matched to an ozone monitor




which reported  fewer  than 168 of  the 336 possible  hourly readings during




the two-week period (24 hours x 14 days).  The following results, then, are




predicated on  the  air pollution monitors nearest  the individuals' houses,




all within twenty miles, as well as on thorough reporting at each monitor.








     4.2.1  Adults




     The  first  results  to   be  presented  concern   the  total  number  of




restricted activity days  (TRADs)   due  to respiratory disease  during the




reference period.   This variable differs somewhat from the MRADs variable




used  in  the   "all  causes"  analysis.   There  attention  was  centered  on




individuals who reported RADs but no work loss or bed  disability days so we




could be sure RADs  were "minor."   In the present analysis RADs may include




work  or  bed  disability  due to  respiratory  illness.   This  alternative




approach  was   adopted   because  relatively   few  individuals  reported  a




respiratory illness that  did not  result  in work loss  or  bed  disability



(both of which are analyzed separately below).




     Table 4-7  presents the  analysis of total restricted activity days in




adults due to respiratory illness.  The general approach is that adopted in



the "all  causes"  analysis.  First, simple ordinary  least squares was used




with  several   different combinations  of  air pollution  variables.   Then




variants of this plus other approaches were employed.

-------
                                    4-32
Table 4-7.  Restricted Activity Days Among Adults:  Respiratory Disease
Model #
(48)
(49)
(50)
(51)
(52)
(53)
   Dep. Var.
Indep. Var.
                                                          RADRSPPS   RADRSPPS
              TRADRSP    TRADRSP    TRADRSP    TRADRSP    (LOGIT)    (LOGIT)
INTERCEPT
03NR01
03NR01SQ
S4NR01
SPNR01
N2NR01
CONE01
RACEW1BO
SEXM1FO
MARY1NO
INCOMCON
FAT
FATSQ
AGE
.724
(1.95)
1.544
(1.98)
~
-.00334
(1.08)
—
—
—
.103
(1.99)
.0403
(1.10)
-.0177
(0.47)
-.00000470
(2.37)
-.605
(2.32)
.120
(2.39)
.00981
(1.77)
.536
(1.44)
1.873
(2.32)
—
-.00154
(0.48)
-.000302
(0.53)
—
—
.100
(1.94)
.0384
(1.04)
-.0186
(0.49)
-.00000465
(2.34)
-.601
(2.30)
.119
(2.37)
.0101
(1.82)
.791
(1.75)
1.409
(1.46)
—
-.00255
(0.61)
-.000594
(0.84)
.000671
(1.48)
.00476
(0.64)
.108
(1.75)
.0430
(0.96)
.0167
(0.36)
-.00000627
(2.62)
-.795
(2.51)
.161
(2.63)
.0105
(1.58)
.561
(1.51)
—
7.479
(1.75)
-.00128
(0.40)
-.000235
(0.41)
—
—
.103
(2.00)
.0389
(1.06)
-.0165
(0.43)
-.00000465
(2.34)
-.600
(2.30)
.119
(2.37)
.00996
(1.80)
-3.939
(2.61)
6.231
(1.79)
—
.000578
(0.85)
-.00234
(0.00)
—
—
.454
(1.77)
-.126
(0.75)
-.352
(2.10)
-.0000126
(1.40)
-.951
(0.99)
.175
(0.97)
.0408
(1.55)
-3.972
(2.41)
3.650
(0.92)
—
.000525
(0.76)
-.00241
(0.00)
.00172
(0.96)
.0168
(0.54)
.379
(1.37)
-.106
(0.57)
-.247
(1.31)
-.0000239
(2.28)
-1.272
(1.26)
.239
(1.29)
.0601
(2.02)

-------
4-33
Table 4-7 (continued)
Model #
Dep. Var.
Indep. Var.
AGESQ
SMOKY 1 NO
EDCOMCON
CHRLMDUM
AVMAXTMP
DMINTEMP
DMAXTEMP
AVPRECIP
HUMIDRF
R2
F or X2
N
(48)
TRADRSP
-.000111
(1.89)
.0371
(1.07)
.00206
(0.36)
.247
(5.29)
-.00401
(3.26)
—
—
.119
(0.70)
.00190
(0.9D
.0142
4.41
4,906
Restricted Activity Days Among Adults: Respiratory Disc
(49)
TRADRSP
-.000113
(1.94)
.0364
(1.04)
.00177
(0.3D
.247
(5.29)
—
-.00138
(0.83)
-.00265
(2.67)
.128
(0.73)
.00243
(1.17)
.0150
4.14
4,899
(50)
TRADRSP
-.000112
(1.59)
.0466
(1.10)
-.00121
(0.18)
.256
(4.51)
—
-.00233
(1.14)
-.00316
(2.61)
-.000492
(0.00)
.00228
(0.86)
.0190
3.56
3,703
(51)
TRADRSP
-.000112
(1.92)
.0361
(1.03)
.00175
(0.3D
.245
(5.25)
—
-.000669
(0.42)
-.00193
(2.16)
.104
(0.60)
.00231
(1.11)
.0146
4.00
4,899
(52)
RADRSPPS
(LOGIT)
-.000553
(1.92)
.255
(1.67)
.00903
(0.35)
.565
(2.98)
—
-.00514
(0.79)
-.0146
(3.38)
.275
(0.36)
.0250
(2.50)
73.88
4,899
(53)
RADRSPPS
(LOGIT)
-.000701
(2.19)
.219
(1.25)
.0130
(0.45)
.457
(2.14)
—
-.00765
(1.01)
-.0142
(2.9D
-.0914
(0.10)
.0264
(2.28)
57.11
3,703

-------
                                   4-34
     In equations  (48) -  (50),  ozone was positively  related  to the number




of  RADs  due to  respiratory disease.   In the  first  two  regressions  this




effect was  significant at  the  5 percent level;  in the third,  where  five




pollutants  are  included,  the  coefficient was  insignificant.   Income was




negatively and significantly related to respiratory RADs.  So  was average




maximum temperature  in equation  (48)  and DMAXTEMP  in equations  (49) and




(50).   (The  latter  measures   average  daily  maximum  temperature if the




respondent  was  interviewed during  the  second  and third  quarters of the




year, and equals zero if the respondent was interviewed during the first or




fourth  quarters.     This  distinction  was   introduced   because  higher




temperatures in  the  winter may  be  good while  those  in the  summer may be




harmful.)     Both  FAT  and FATSQ   entered   the  equation  significantly,



suggesting that there  is  some ideal  weight-to-height  ratio which minimizes




respiratory  RADs.     Whites had more  respiratory  RADs  than  blacks,  a




difference that was  quite significant.   The chronic  limitation dummy once




again  was significantly  associated with  acute  morbidity as  measured by




respiratory RADs.




     As in the "all  causes" analysis, we  experimented with a squared  ozone




term.   In general,  it  did  not  outperform  the  simple  linear  measure,,



Equation  (51)  represents  an illustrative  case.   The square of  ozone was




significant  at the   10  percent level  as  contrasted  with  the  otherwise




identical  equation  (49)  where  the  linear  ozone  term  was significant at




above  the  5 percent  level.   Equations  (52)  and (53)  are logit  analyses of




the  likelihood that  an individual  will have at least  one respiratory RAD.




They are  consistent  with the OLS  results although the  ozone variable was

-------
                                   4-35
somewhat less significant in the logit  analysis  than in OLS.  As expected,



most  variables  exhibit  similar  effects,  although  FAT  and  FATSQ  were no



longer significant in the logistic framework.



     The performance of  the dummy  variable for smoking is worth mentioning



here.   Even though the  focus  in  equations (48) -  (53)  was  on respiratory



disease, the  smoking  dummy was generally less significant  than  it was in



the all-causes regressions.  Although its sign was always positive, only in



equation  (52)  was  It  significant  and  even there  only  at  the  10 percent



level.   One  possible  explanation for  this may  be that  individuals with



chronic respiratory illnesses, who would account for a substantial share of



acute respiratory illness, have quit  smoking on their own or on a doctor's



instructions.  Thus, they would not have been identified as  current smokers
                                        •


in  the  HIS  even   though  they once  smoked.    In   subsequent  sensitivit'y



analysis, we probe more deeply into this possibility.



     The results from the analysis of work loss due to respiratory disease



are  presented  in Table  4-8.   Equations   (54)  and  (55)  are OLS  estimates



using ozone with two  combinations  of other pollutants.   None of  the air



pollutants  was   significant  in  either   regression  with  the  exception of



particulates  (at  the  10 percent level)  which  had an  unexpected  sign.  In



(56) and  (57) the  square of the average  daily maximum ozone concentration



was used.   This  substitution changed  the  results  little, if at all.  Among



the  other  explanatory variables only education was  significant  at  the  5



percent  level—the  sign indicating  that  individuals  with  more  education



have fewer work loss days due to respiratory disease.  Even  the presence of

-------
4-36

































Ul
-
0.
V
>.
i-
O
4-*
f*
L
"*
tfl
i
nc


irt
yi
c
«j

Jt
C
o
T

*•*
»•«
3
-o
**

00
)
%f

It
rH
.O
n>
*—

3
w






•o
w/





^
i«
W





.*•'»
r>
UM
w




[7




-»^
^•'
U~i
w





•-»
l/>
>./






^
^
u>
^







*

-^

•^
o
5-
a. »*%
d. 1™
CO |-4
« O
-i a
IA ^j



l/l
a. ^
a h-
QC O
*.' C.3
WO J
3 W



5;
—
*/i
tT
H-


a
w»
or
_J
vi
3


O.
O.
^


o.
cfl
Cr
fj
l/l
"X
*—


Q.

-J
fc/1
^»
H-




Q.
CO
Of
_J

.Z
»—


L •
> 13
>
•
0. •
01 &
O 1»
•»?
C


• • • • o *
fO ^ ^ O i ^




vr /^ •-* r> i &• ***
rvj ® 00 P* 1 -O (^
. . • . O *
/\j — — O • O
1 w I w I w




0 w-i ^ 0 | ^ W
— * O * 0 •
o • o o o
W W 1 W
1


^ ^ ^ ^ , ^ ^
On rsj O •*• 1 *~-j m
w* Cfl ^* <"-J O ^^
f • t • O .
0 O 00
w w O w
t
1

f* • » » O •
f O ^-» O O O
1



-4 O t ;n — — 1^.
-.4 i/^ ^ r^ ** in
• » • * O •
 m >»•
i— n a v»

o « o • o •
o o o ^» • o
1



03 -^ ( t
*> O 1 1
QD O
O •
0 O
O w
o
f
I

f* t.- fr O r--: o
0 t O • 0 «
O — i O <-« O O
C» v^ O w « v-»*
* •
1

*. '-V 1 |
r* O t t
>f ~f
0 •
o -*
0 N>
•
1

r^ « o • o •
o -« o — o c
• t
1


rg •« | 1
O •
O P«
O w
•
1


o ^ -o •* ^ ^
O • r-» t O •
O •-* O « O 0
O ««/ O *-» • ^-«
1 »
1



^ ^ 1 1
^f\ 0s 1 t
** r\i
0 •
0 —
o ^
1
1





f-*l «-4 «^
o o o
a: cc a
z r z
a. *N a
u^ Z 0

o» «• •• o» ••
• O 10 |O OO « i-*
o
o
t

O »^ lA -^ i^ ^ f* *-^ fii^ ^
-•f* «^ en-^- »^sf CTCP
• • Oi • • O* ••
0 • O 1 O O O — <-*
w 1 w w O ^ 1 ^
o
0
•
t
r^e" **O sTi/" fvj*P ^-">
O • O • O • O • O •
OO • O OO OO *O
• V,/ I*-* »\^ O^X ts-*
> 1 O
0
1
f 1 ^ IT ^ O ••* f" ^ '^" "*^
g^ r\j O OO ff' *«* | O 3 i*^
f i —^ vT »-^ O^" I1 ' CC **• -ft
Oi O« O« O» Oi
O O O O O O **4  o* o»
OO • O to OO • O
f ^> 1 ^-* t >^ O w ^
o
0
1
1

.* r— lA .-i \Of< 1 O rjsf
--<** (T-nt ^-«f». LL,CT r-} ^
O « O • 0 . — . O*
• O OO »O COO »O
-*^ • *-w 1 w o ^ t 1«^
r •
^
1


o z
•« o o n
^* ^ 2; (j
3 9« *+ 21
LXJ 2: >- a
w x a o »-
*l UJ < S <
Cf 



,. ^ _ ^
-< >r a. o
^4 !*•> r r».
O • 0 t
• o o o
^ » ^^


r* , ^-, p ,-S
0 C? CO
I



O T> ^^ f
0 . 0 •
T1 C* O f
* W t W





•3 . 0 •
O 0 O O
» w • ^
1




.-« o -< a.
n- V-H .» **•
Of 0 •
o o o c~>
• *-^ • <*s








(*9
U1
hb u<

-------
4-37



















*l
1,1

J
~
>
«
^
r-
l_
-«
^
1-
'•>


.,

l/»
o
_J
-*
O
-5

**
—4
3
77
»^
,
TJ
3
C
"Z
C
0
u

1

n
-*
'0
^

^
*•«
^
-^



••>
•o
^


l^



~
..<
w




^

*_,






,^
o
in
Vj/





--N

cc
^J
^1
75
h-


'V
-J
•SI
^_


L. •
"3 U
> 0)
=>
a •
D a
CJ 4*
C
*-*

OO ^,f 0^- xT-r O^l A| IT* 2»* ^» ^ s*
r*^* t^>^ U>CO •* *-t
o
1
rg ^ + * * ~ ,0^% ^^ 0^ ra^ <^^ ^^
tf^^o MOO (MO ^r» o» (no* r^oo mm -M»-
Ot • » O * *t Ot O* tt Ot tt
O O 1 O O O O O O t >•* | O • ^i Q
• 1
1
oo oo oo oo oo oo « o oo • —
Oil O t •
• • 1
1
0»n 0-0 CT* T* M»- U>^ f*m ^ r^ -40 ^t-
O« Oi Ot O» O • O» O» O« Ot
oa t»^ oo »o oa oa «o oo t-*
Ow is-' •'W ^^ O W O*-/ I'w *\-/ »»^
^ t « •
* I
P«^ ^.^ ^^ 0^ a» /^ c^^ -^^ 1*1 ^
h-O ^ui o-i r* »• ^ -* cn^ u^CT* .^'A ,
o« o* o» o» "3* o« ot a»
OO «O «rj OO OO OO «O O**
O\-> t'^' tw »^^ O^/ 'O^-' l^^ «w
o lot
« O t
1 t
1

u^ ~. oo >»• ^ jj -. so « -n -o r* NO ^ -ts 1
irt^H >* j •*! -^ C »* Q t^j ^1 f "33 ,f ~f v* u>
O« O» O» O* O* Ot Q» O«
O "^ » "^ O ^J ^ O f^ O O "3 • O O ~*
Q ^^ 1 *-> t^/ O ^^ O^^ O >-* 1 ^* • -^^
0 t • • t
• 1 1
I
^^ rj/-N -0^ ^/-» O*/-* ^1^ 0/-« -4- * |
h-O ^»ir> OO fs f» r«.rij v^f^-i r— <*i 330 i
lAO^ <^^C r^fn f- .-* fM O u«h* r-4 *^ t-4 — t
o« o» o» o« o* o« o« o»
OO to • r*4 OO OO OO »O O*^
Ov*1 )v-/ Iv-* »**^ O ^ O^-1 |*s/ t s^
O IO*
• * 1
1 1
O • Ot Ot Ot O* O* O • Ot
0*^ •-* O*J O« OO OO »O O-^
0 | t • t
t 1 1
1


o z z: a. a. a.
t-« 32
 r> 1A
• 1*4



r- o r~
o • 
-------
                                   4-38
a  chronic  illness  or  limitation  does not  play  a  significant  role  in




explaining respiratory work loss.




     Recalling the  effect  that a  blue  collar dummy had  in the all-causes




work loss equations,  several  models were estimated testing its effects on




respiratory work loss.  Two of  these  regressions  are presented in (58) and




(59) in Table  4-8.   It had the anticipated sign  in both  equations and was




significant at  the  10  percent  level  in  (59).   Again, this  suggests that



blue collar jobs might have health  or accident  risks  that  affect those




working at  them.   In these models,  education was no  longer significant,




possibly  due  to collinearity between education and occupation.   Finally,




the  sensitivity  of   these  OLS  results  was   explored  using  logistic




regression.   Equations  (60) and  (61) present the  results  of logit models



which include  the  blue collar  dummy  variable.  The analysis indicates no




significant effect of any of the air pollution variables.




     Before leaving the analysis  of work  loss  due  to  acute respiratory




disease,  it is important to comment on  the  F-statistics of the regressions




presented.  They were low.  In none  of the regressions  can  one accept at




the 5 percent level the hypothesis that the entire set of  regressors (with




the exception of the  constant  term) has any effect in explaining variation




in work loss  due to respiratory causes.  These are the only results of all




those analyzed for  acute  illness where this  is the case.  The problem may




be due in part to the relative infrequency of respiratory related work loss




in  our  sample.   The  average  number  of respiratory-related work loss days




during  the  two-week  reference  period  was   0.05   (in   other  words,  one



respiratory work loss day every forty weeks).  In the logistic analysis, of

-------
                                   4-39
the 3,000 or  so  individuals included in the  different  runs,  only 50 to 60




had at least one respiratory work loss day during the reference period.




     Table 4-9 shifts attention to  the  analysis  of bed disability days due




to acute respiratory  disease.   In keeping with  the usual custom the first




three regressions,  equations  (62)  -  (64),  present  ordinary  least squares




estimates in  which  the  average daily maximum  ozone concentration over the




two  week  period   was   entered   linearly   in  the  company  of  different




combinations  of  air pollutants and other explanatory  variables.   A fourth




equation, number (65),  used  the  square of  ozone  to explore  the kinds of




non-linear effects suggested by the earlier analysis of restricted activity




days due to  all  causes.  While always  positive, ozone was not significant



in these four bed  disability  regressions.   Neither were  any  of  the other



air  pollutants  with  the  exception of  nitrogen  dioxide  in  equation   (64)




where it was  positively associated with bed  disability and significant at




the 10 percent level.




     Among the non-pollution variables, income,  general physical condition




(FAT),  maximum   temperature  (in  the  summer), and the  chronic  limitation




dummy were also  significant—as they  have  been in many of the other models



investigated  to  this point.   Age  and its  square moved in  and  out of




significance  depending on the  particular specification of the model.  Note




that the  present smoking status  of those  in  the  RFF  sample,  while always




positive, was never significant.




     Equations  (66)  -  (69)  contain  the  results  of logistic  analysis of




adult bed disability  due to respiratory disease,  using both  ozone and its




square.  The  results  did not differ much from those obtained using ordinary

-------
4-40























a*
n
.-
O
..,
0

X
t_
o
*rf
L»
•4
a
t«
&
IV

• •
>.
•«4
*4
-O
1*1
i/l
.*rf
Ci

•a
*
Oi

*rf
i-4
3
17
<


•
m
1
^f

O
«-4
,a
fQ
*-


O*
SO
w


/•N
sO
W



p*
O
w


/-s
O
«o
w




*"*
itl






»r
•0
vx



^
**'
•o
w







/^
^J
H>
W








^

•H
*

o
Ti

CL *-v
L-1 *-•
oc o
o o
UJ «J
CO w

a. /^
a. h*
Ot O
a a
UJ J

a. *•»
t/> l-i
ne LT
O O

(C w
lS>
CL **•
Q. t-
oc o
C O
u- _j
(i- \s

C.
L/1
at
UJ
-*^
J—


CL
C"
Q
U'
^»


i/l
oe
C-
UJ
n"
M.




Q.
in
f
O
u-
(D
H-

«
t- •
a L.
> "i
>
t
a •
* a
o *
•o
c
b<4

^* r* ( -o <& r» r—
• t t • o »
f** ^ O O • O



O» ^ 1 «^ in »- in
-* fi »• CT- >o «/
1

fl 0 -J- r. 1 !C -
* t • * O *
rl ^* — O • O



• • • t O •
sf Aj ^- O O O
IN-* W » -M*
(


r\j x^ j so ^> CS ^
rv. I/' | «J- O r-- rj
<0f *"" *— m NT O
• * « 4 f"1 f
<^, O O O
\^ s^ O v-»
O
•
1

• • O • 0 *
o
o
1
**J 4 *f | )
^i j> —• O
• * O *
^^ • C3
W S^«





k-
0. O
1_J l^
^J ^^ *« fM
(V O CT O
yj ft QC Qc
»- J 2 3T
2 *n m .#
<-i a o L^

^ rg
O i
O O
•


(SJ (^
O •
0 0
• v^


C* •*
o •
o o



o •
O 0
• s^



UT ^«.
N. fKJ
•^ «"l
o o
O s^
•



0 •
,
1

-* 3-
o •
o o
o w
•




(•* *^
f*1 ^
''J O
o •
O — t
O w
•







*^
Q
QC
z
a.
Lrt

f1^ 03 O ^ O^ N> ^^ O O i^J t/^ ^^
O* O* O* •» •• O •
O i— «O • O IO 1*4 O~4
0
t
1
»-i u"> rvj r^ i/t wn v^iO
O t •-• 1 rw O*^
w s^ \^ O v^
O
1
fiu^ O^ CT^- *-O O'J tr» ••*
o » o* o » •<• •• o*
o*- »o • o if1 I— « rr^
o
•
1
1 | mir O^ **•  -^-^ Cf \fJ •^<.
rvjcj m^ r-^ri «*-i^
o» o« o» cn»
• O ««^ »*•« O r-u
h*- \^ | w O ^>
o
o
«
1

1 | ^ ^ i/> ^ <^ ^ O O
1 f «^cn-Of*Ot'1'*O'O
r-j (?» i*-! (^ T* -j "MO
o* o • o* o .
• o o o o o o f**
*w* e \^ • \s O v^r
1 1 0
o
1



o z
. c
2: ^ l_> X QC tJ
P»J a «* ui « 2
2 o a: trt r: *-i

u. m ^ ^ a* -
• • » . o •
•H ** «t • I^J
'


v4 *N rv4 (\J W"» CP
• t « • C '
1 *•* *-/ v


ui n >» C" Os r-'
• • • • 0 •

t w


rj O ft tf w> ^»
— 1 — M * —I
I ^* «^ »^



(~\4 -^ ' •* '~* -J- *^
U" f UP' (-- .
r- s »— i— w: c
1 vf t^. ^ f-J
V> «^' * •— '




— . i • O •
. .r- (r 0 —
'


>- -c *-. ^- ,- cj
. • • * er *
i *j tr O 'J
**-* w * \^>





/«-» /^ rn *<+ o *^
•-^ C^ O ^- «*J fM
ir «^ «• L,* ^-j r*>
* * • • 0 •
| %y vf C5 f~*
^^ ^X • S^









C»
oO
W- ^- Jj
< 
-------
4-41







































/•*
•o

ft
&
10
*—
cr
w



0?
*o
w



r-
M




w»
*£>
s*»



*








£
w





^
CJ
>O
s./










V

*-•
«
-o
0
c
0. »»
4/1 O
cc c
U' ^

a. ^
a. —
w< O
C£ O
0 -J
CO
a. >•-
C. h*

or o
a -j
IL' W

a »-
rx M
i/O C1
o; a
a j
c
Q.
Ct
rs


«
O
(U
a<
*-

VI
o
LU
rr
^


^
(XI
DC
C3
itj
03
fc-


•
L •
"9 U
> T)
>
•
a •

O i/>
O •
O t\t
t
l/N *-*
ce ^
Ov ^f>
o »
O rg
0 w
I
0 Mj
O «
O M
• v^
t

CT* *f
CT* ^
o *
O *M
O w

^» ^
O •
O w
O
*
1
<* e-
o •
o —
0 W
0
1
^ tn
0 •
o ^
O v^
o
•
1
as ^N
97 «f
r\i ^
O •
O -*
O w
O
*
1







c»
l/>
IU
L5
^
s^; °* ^^ s« ^c *s ss
o* o • • • o • •• o* ** •
• O • O ^* f ^ O • *M *fl ^"»

h.0-
&• *Q rn^rf rn i/> <-sj r*- rj *t c^ cc O9^
O <*i rg f^ ** P"* ^« ff* «rfv4 frto *TI
e» • O * t* a* t* o* LA *
• O 9 O <** * ^n O • ro cc "J"
w | w w 1 w w w

£S ^P ££ ^2 S^S *7 ® S
o* o* •• o • • * o* *?»
• O • O •* i ^ O • ^J ^ ^^
w j *»/ w | w >^ v-f

o.«
^n\T nr*^* (*^x/- r^irv IT'ST CT"Wi OC^
«O ri r>* f— sr t~ r-* ^- *•**<* m w- * a:
o » o • i • o • ••o* *<^*
• O »O ^* t *n O »r« Qjj-
W | W W 1 W W W

• 0 OO •<"* Oi^ »O 0— • •*
t •
1 1

^ CD IS, ^ (* rM Om ^^ 0*^ 0«0
o» o« ^» o» o» o» ^^i*
• O O """3 • '"•J O *^ O C ^ O t l*^
^/ • *•/ ^»rf • s^ i w/ O *w'
1 II.

M «%. /nm m^ ^ (•* 0 ^ tf- ^ ^^J^
O • O* Ot O« Oi C* t C1-*"
« O OO fCSJ O'N OO O «~< • "X
W O^> ^"^ O*-' »s^ •*«'
• * 1
1 1

.0^ o,-v ^^ CT^N /o ^ C*^ f-^rrj
PTIM O<^< J*-CF> *O^ mO »^(*> rKODCZl
(^ vf> 'A O ^ It? ^ ff, ^ IT. •« »f O t *"*")
O • O* O* O* O • O * O NT •
O^ OO t **> O"J «O O-^ • **
fly^Ow^ v>O^/ v^ t v^
o •
* i
1




o as x a. a.
Z C "^ Jf M 11. I*J
»^ l_) O UJ O tt X
>- IL f <- UJ 0
K; a -i z oc i-t u
a u n: « d- z o
x o i r > zi 
-------
                                   4-42
least squares.  None of  the  air  pollution variables was significant at the




5  percent level  in  the logit  specifications.   The  dummy  variable  for




marital status was  significant  in the logit specification  although it was




not in OLS.  It indicates that married persons are less likely than singles




to  experience  respiratory bed  disability  days.    The  other  results  were




similar to those obtained in OLS as well.




     In  reviewing the  analysis   of  acute  morbidity among  adults  due  to




respiratory disease, one conclusion  emerges.   Although  there appears to be




no  connection between average daily  maximum 1-hour concentrations of ozone




and either work loss or  bed  disability,  the results suggest that there may




be  a positive and statistically  significant relationship between ozone and




total restricted  activity  days  due to respiratory  disease.   In two of the




regressions in Table 4-7 the coefficient  on ozone was significant at the 5




percent level and in two other  cases  it  was  significant at the 10 percent




level.  It is important to examine this association in more detail.




     A  considerable  number  of sensitivity analyses were performed  on the




association between the average daily maximum  1-hour concentration of ozone




and acute respiratory morbidity in the form of restricted activity days.  A




number of these results are reported in Table  4-10.




     To   begin  with,  we  considered  the  possibility  that  imposing  a




requirement for at least 168 hours of valid ozone data during the reference




period may have  biased the results  (this  might  be the case, for instance,




if  the  monitors   in  the  highest  ozone areas  were the ones  most  likely to




have  complete data).   In equation  (70)  we  duplicated  the model  in (49)




(Table 4-7) but included the 400 individuals whose ozone data did not meet

-------
4-43












0>
VI
(C
*
(A
Q
^
^
+t
i*
L.
w*
Cl
*
QC

*.
X
**
,*4
.*
o
•o
(t
••4
L
~
a/
oe
^
•^
3
13

f^
rtl
C
**
>
•4
•H
tA
C
*
'/>


•
O
V*
1
.«•

a-
•4
A
ro
>—


*^x
r*«
t*>
w





^-n,
W^>
IN.
v^




^%
u^
^fc
v^





/•^
^J.
»0
w



m






*^
nj
f-
v^





r«s
_
^^
W



O
f*.
v^








sr-

-^

T^
Q
IT

a
1/1
QC
o
«r
&r
t-


a.
IS)
ff
a

- CJ

a.
2
r»
3-1
V-


a
LSI
Cf
o
^
QC
K*


c.
^J^
ry
^;
^
fV


O.
to
ft
O

•
M. •
It Q.
Lj a»
T)
C
*•*
O -*N t t t | m ^ ii
i-« CO |t II *O -* II
M »- f* O
* • o •
O • 0
w 1 \^



m ^s it &* <** II II
'MO* li ru o ii II
rg *y O O*
t • ft
I O <-» O
w w



o ff* o* *- ii ii o o
Irt ^« fT* fi ^* ^
• * • • o *
wrf »*  (N* • *•
1 V^ — i W t W


••* ^> ^ /> || |) fNj ^
tA fn L^ f t 1 it •"• r^(
4 • t • O *
og fsi O O O
^^ *-/ • w
1


^ ^ m ^ l , , , p, r%
•* «" r- — || l| s» T-
l^ ••* ^* *j- ~* r*^
• t • • 0 •
*^ »* ^ o o
Vrf ^-> t W
1


C* •*% o ^ ii 11 in *^
IA %o ^* •*% || ll r~ N*
^» f«- m m rg o
• * • • Q •
»* •* u^ rg O O
y^ *^ O ^/
I

»J ^ ^> ^ f 1 II ^? ^
^ — (M ^ || || h- r-.
• • * • O t
•4 f-i ftJ O O
W w f W
1




^
O. — -^
UJ O O
U •* QC OC *-
rz C3 2 2 O
UJ ct O tn Q;
^— 2 -^ O 2
2 en r>> rn *t
•-* c: a 3 \s>

|| O»O 1| It IAW% vff* O ^ O
o* o • o » o •
* ^ o o • o • —
O W 0 w w w
1 •


co/^ II II «--% li in^ -O^
^riO II II »f*^ II ini^ o^
^ o mm ^- -o sr «
o t a* • t o »
o o o o <\j « o
O v' O ^ ^ ^
O •
• f

il tl eo» II it -« n o *-
^4 *^ »-^ v« >J <^
O • • t O •
O O <-a » -*
o ^ \»* ^^
•
1


t 1 |l r**^ |t II S»0 ^>0
•-» fNJ C^ *C ^* O
O • t • ^ •
O 0 — • O
* w w 1 *^
1

ii ii ^ || t| vtO- r^r-
u~\ in ^« o 3-' •"•
a • • • r> *
00 — 0 0
o *•> •»-/ • ^^


II |l irr-y || ii oi-w tfiivi
O • • i O •
O O r-J • O
O w *^ ^
«
1




„ — M — 0
O o O O pn O
Q; QC t-< QC QC .* U-
2 2 O 2 2 3 •—
o t^ ot o in is t TL.
^ 0 2 r« O L.. X
-r -» a. Q. o. «» iJ
ist ».•» ty? u^ i^> or tsi

p* a> ,^ i-f
*NJ rrl ^1 -f
o • o •
» O O O
I W 0 w
O
0
1
« ~ >o •->
»n <** '^1(7*

O • O •
• O O M
1 ^ O *•*
e>
o
i
vo r-. t- m
^4 x* I/ n
O • O •
• T* O r '
t w O ^
<*>
O
I

v- vr .0 *-
ng f>- — •> xf
• • r~- •
1O O -•
•fcrf C' N-*
o
I
f. -^ /^
0 - C=
^J ,^ f.^
!^^ £1 ^t
^
1

fv: * ,. *
rq ^t /-n r»-
^- 1A f*^ H .
o • o •
» — r"* I—
t *^ u> ^-*
C""
^
1
>!• x> a? ^
P^- "*« T~ *-
rr- „ a r-
O • fi »
• o c. —
w 0 v>
^
1
•^ O sT O
O • O •
0 O 0 ^i
* sy O '»'
1 O
O
t



^
o *t
2 LJ
-^ £
^ CJ
(V U1
< ^
E H*

-------
4-44





























"*3
31
C
-*
w
C
o
-*
1/1 !_;
ac a
•s*
Q.
^_

a
^
h—

L/1
Qt
O
or
H-

'X
a
ac
H-


* J
>
•
a •
» a
O 1»
-3
c
M

«j ^ ^^i mr^ r^rM
O • O* O « O •
oo to oo oo
1 O

O t O* O • Ot
t O • O » <*g O *M
t
t
»/) if ** US "^ U> ••• «O
• • tt Ot O *
1 **l CM O '^ O •>*
s^ N^ • w O w
r
_^ ^^ ^^ ^^
-f -^ *n '*^ en -M m •*
v/i^ *i»^ src^ mo
*• •• o» o*
-* o »•« • o o *•»
'*
* * •* O« Of
I w s^  U%U^ (M0N (M(N OP* II 0*l/>
O • O * * » O * O * ••
»OOO ^ • S^ ^>
1

^ 04 -O O p<4 ^ rj o *4j fJ
O**Ot»« O*O»
r* • V> **> • s_/ | w
>-» 1 I

mo h-~ ^n* ^^o »n m it -» *
O*O* ttotO* tt
V>*S^ %M^ | \^ • \^« \^
t



o 2 x a. a. a. a.
2 a 3 T: s: c —
i-t l_) Q uu JJ h- tj
^ r x — »- x uj
atf a -* z x • >
*s> uj o a a «* <*

o r*
•^1 IA
O t
o o
t w

O t
0 0
t w

f\* y
O •
O O
« w
O -*


^ ^
^ ,^
<-M T-
O •
O 0
t w
1

1 «
1 -^
* w


O i—
O *^l
t t
*^
\^





u.
ct
o

r
3
i

I I | 1 *f C^ ^
II II ^ ^- -*
0 0 t
t •*

t I ||  (M <7»
II It ^ -t f*.
0 t ~
*n


^-* • ^r
s>

II II f* -O
II t 1 ^ ^
* OJ
l^i
rlj

CO *^ II »^ ^4 >/
un ^o ii ^j <*j -«
^j >O m • (*i
O • C" -f •
O 0 t ^i
» v^


II 1 1 '>J o ^*"
II ii n ^» o^
<-t » r
o I-HJ •
• *j



|l I) !"O v3 -*T
t 1 (I O r*- ^
f^l • *^i
o ™ »
« OJ



1 1 II *f 'f* *f
| 1 II >* ~ ^0
o -*• »
» >ri





u_
nc o "^
22 X
UJ *-4
-1 » t-
_J V- 0
a <-o ^4
x a. ae a. s

-------
                                   4-45
the  168  hour  data completeness  criterion.   The  results were  virtually




identical between the two regressions.




     Because the  period between  April  and October  each year is  the time




when ozone  readings are  typically high, a  number of  runs  were performed




dividing the sample in half.  The  purpose  was  to determine whether it was




summer exposures  alone  that  were  responsible  for the observed relationship




between ozone  and respiratory RADs.    Separate  models were  estimated for




those  interviewed  in  the first  and fourth  quarters  of  1979, and those




interviewed in the  second and third quarters.   These results are presented




in  equations  (71)  and  (72),  respectively,  where ozone was  entered with




sulfates and total suspended participates.  In spite of the fact that other




coefficients varied substantially between the two subsamples (participates,




the marriage  dummy, and FAT  and  FATSQ, for instance), the  coefficient on




ozone was virtually identical  between the  two.   As in  equation (49), where




the  two  subsamples  were pooled,  ozone was  positively and  significantly




associated with respiratory RADs at the 5 percent level in both subsamples.




     Next, we  examined  the sensitivity  of  the apparent association between




RADs  and ozone   to  the  inclusion of  a measure of  pollen  concentration




(POLLENRF).  In equation  (73)  the relationship  estimated in (49) (in Table




4-7)  is  reproduced for individuals for whom  pollen  data  were  available.




Including pollen  in the regression reduced the magnitude of the coefficient




on  ozone  by about  half and  reduced its t-ratio  even  more.   This  suggests



that effects apparently attributable to ozone (or other air pollutants) may




actually be due to  allergic reactions to ragweed  and other pollen.

-------
                                   4-46
     Because this would be an important conclusion, it was pursued further.



Note that the "without-pollen"  respiratory RAD regression in equation  (49)



is  based on nearly  4,900  observations.    The  "with-pollen"  regression




presented in  (73)  is  based on  2,313 observations, less than half as many.



(This difference arises because pollen data are only available for a subset



of  SMSAs.)    This  suggests the  possibility that  the  change in  the ozone



coefficient  may  have been  due  to different  samples  rather  than  to a



spurious correlation when pollen was omitted from  the estimating equation.



     To  control  for the heterogeneity  of the samples,  a  simple procedure



was  employed.   Using the  same  2,313 individuals  upon  whom  the results in



(73) were based,  another regression was  run  using respiratory  RADs as  the



dependent variable but  excluding  pollen as  a regressor.   The results of




that test  (not  reported here)  lend credence  to  the  view that  it  was  the




change in sample  size  rather  than  the  inclusion of pollen that reduced  the



size and significance of the  ozone coefficient.   In fact, when pollen  data



were added  to the  respiratory  RAD  regression on  the  smaller  sample,  the



coefficient  on  ozone  increased slightly in  both size  and significance.



Thus, the estimated coefficient on ozone may be more robust with respect to



pollen concentrations than the results in equation  (73) would suggest.



     Besides  children,  another group  potentially more  sensitive  to ozone




and  other  air  pollutants  are  those with  chronic illnesses,  particularly



chronic  respiratory disease.     For this reason,  866 individuals suffering



from any type of chronic illness  were  identified for additional analysis.



RADs due to respiratory disease  among this  subsample  were  analyzed using



logit  as  in equation  (52)  above.    The results  of  this analysis   are

-------
                                   4-47
presented in equation  (74).   The coefficient on  ozone  was  more than twice




as  large  in  the  "sensitive  population"  regression  as   in  the  "normal




population" model and it was  even more  significant  (at  the  2 as opposed to




the 7 percent level).  This finding is  consistent with the  hypothesis that




chronic  illness  may   increase  one's   predisposition  to  ozone-related




respiratory restricted activity days.




     Earlier in this chapter it was hypothesized that information about air




quality  conditions  might influence  estimated dose-response relationships.




Such  information  might  alert  people  to  poor  air  quality,  with  the




consequence that they  avoid it (and thus make  it difficult to observe any




adverse  effects  of  air  pollution  on health);  or,  it  might  actually make




them aware  of  minor ailments they  might  have  otherwise overlooked (thus



strengthening  the   observed  relationship   between   air   pollution  and




morbidity).  As  a  test of this hypothesis,  we  included the dummy variable




PSIY1N0  in  a  regression  explaining  restricted  activity  days  due  to




respiratory disease.  As noted earlier,  it  takes  on the value unity if air




quality as measured by the Pollutant Standard Index was reported in an SMSA




in  1979, zero if not.



     The results  are reported in  equation  (75),  which can  be  contrasted




with equation (49) in Table 4-7.   Note  that the coefficient on PSIY1N0 was




positive  (there  are  more  respiratory  RADs  in  SMSAs  where  the PSI  was




reported than in those where  it  was not, all other things being equal) and



significant at the 10 percent  level.  This suggests that people may be more




inclined to  report  feeling poorly if they  are aware  that  air  quality is




poor.  We  cannot  reject the  possibility, however,  that the dummy variable

-------
                                   4-48
PSIY1N0 was  picking  up other  unobserved differences  between  SMSAs  that




influence health  but  that  were not  controlled for with  the  regressors we




include.     Ozone  remained   positively   and  significantly   related  to




respiratory RADs in this equation.



     All of the analysis to this point has been based on various samples of




individuals  who  live no  more  than  twenty miles  from the nearest ozone,




sulfate,  particulate and  other  types  of  monitors   for which data  are




required in a particular regression.   It  is  interesting to analyze a group




of  individuals  living even closer to  the nearest air  pollution monitors.




(If individuals stay home all the time, one would want monitoring data from




as  close  to  individuals'  residences as  possible.)    Accordingly,  two



equations were  estimated  using  individuals  living no  more than  ten,  and




then five, miles from the nearest monitor of a particular type.




     The  results  are presented  in  equations   (76) and  (77),  respectively.




Although ozone was significant in this specification of the model using the




twenty mile cut-off  (see equation  (19)),  it  was  not  significant in (76) or



(77).   A  number of  other  variables  significant in the earlier regression




also lost their significance when the more restrictive distance requirement



was imposed.   However,  because of the F-statistic in equation (77), which




imposes the five mile restriction, one cannot  maintain with any reasonable




degree  of  confidence that  the  entire  set  of  regressors  is  capable of




explaining any of the variation in the dependent variable.




     Several additional sensitivity analyses were conducted using different




measures of  ozone during the  two-week reference  period.  They are reported




in  Table  4-11.  Equation  (78)  again  reproduces the basic results presented

-------
4-49








f*
wt
«!
tn
^4
O

^
t.
o
IS
u
•«4
c.
in
a-
ff_


**
>
••4
«*
O
4
TJ
0)
O
u
w
t/>
2»
rc

**
3
*=
H-
o

It*
f'
irt
c
•f

^
+J
.4
>
**
•••4
tn
c
0*
i/1

n*
L.
Q
£

~
I
^

41
^4
Ji
flj
h-


^
u*
CO
)^t





r*
**
CO
w






/•^
/n
en
^




/>
rg
en
w




/•s
* _4
CO
w




,•*,
o
fr
w




>-s
C*»
F*.
^






^>
CO
f»-
s^





•il

^4
c.
TJ
n
£

O-
t/i
Of
0

CC
Q
<
oc
»—



a.
i/>
ne
O
4
IX
I—
i/>
O. ^
a H-
4/) »-M
or cr
0 C
Qt ^y


Q. ^
a. ^>
C^* f-l
CC L?
O O
< -J
p*» \^

CL /-%
Q. ^«
1/1 b-t
ft O
d a
*5 .j
QC s^

a.
1/1
re
Q
*•*
ce




a.
(/^
QC
0
e
or
h-

> ro
>
•
a •
& a.
O "'
•o
c
^rt
F* ^ I fl ^% | | 1 ^ >••»
^ •* ( vr o i t i *f er
«O »•'• *•* O* f» ^"«
o « • * o •
• o o — o o
1 W -• W 0 W
•
1


f— *•* F*. ^» ( ( ( • «fl /%
W. *-» fj *r IM *J-
O • » • O •
* O f^J -* O O
\^ \> » v-»
|



r- <^ o ^ ( i I i «> /^
co o m *o i i i i m **J
sf r-* in> •* *-* m
O • • t Of
• O *V f*t O O
1 >^ *w • w
1

o **»*•» ^ i i > I m *•*
O 03 -r sr I ( ) i o m
fM o o* >*• «^ h-
* • • • O •
n"* ro co ^* » O
s^ w 1 w



^ ^ m ^ { i i i i
O 'i* •*• ""i 1 | 1 1 I
rg O m ^4
* • * •
m  0s ^^
#v f^ t i r i -j^ *j -^ h*
*r o UL M «-< i^
• • f • 0 *
rr^  w « -w
1

H-
& O
UJ Irt
o ^ ^ "n XT ** --
UL O O O O O O
iu rt; ce u ri- _j Q*
•- Z Z Z 2 Z Z
2 1^ ^\  l*> sj
M a a c o a ui
ry -\ ( i i (
— 3s ) 1 | 1
in ^>
o •
0 0
o w
,
1


r> ^ P^ ^ m *x t t
(^ m », r- ^ ^ '
O • O • O •
0 O 0 O O O
O w O w « w
• t
I


f- 'N 1 ( 1 1
^ >> t 1 1 I
U"v f*
O •
O O
O w
•
1
1 | 1 | -J <•%
1 I I t JL> >r
IA *^
O •
• *•
! w



(?- ^ I 1 r^t r* \
•»> ^ 1 t fNJ ^- t
»-» ^C ^j r-
O • • •
0 O 10
• w w
1

W. "> ) 1 | |
•^ CJ | l | t
** V*
o .
O 0
• Sri'
1

e* "^ ) ( i i
f» m l i i t
•"• O
o •
o o
o w
t
t


/n .-N i i | i
IT. >•- | | | 1
»-» rsj
O •
o o
0 W
•
1



«^ »~4 •*
0 O O
ne cc cr vj- ft.
Z Z Z oo uo
Q. rsi O '•n *n
i/> 2 U CJ O
o ^

«^ "O
* «
**
s«/




fVi ^
1^ W>
• •
^i
<->




h


^J ^
«4* -*
*i »
• •
•M
^/



O -^
r— >r
^ er.
v *
*-•
SM'


T> ^
x» r»
s* ^J
f •
-^
^y


%1 ^*
O «-*
.— <~<
• •
r ,
^




-^ /-N
o K-.
•-t O
t ,
*~J
v^



O
5J
^ -
JE
U~
LJ
*T
QC

ir. ^J
fM ^
C? •
• o
w/




V. ^
CP ^»
O •
0 O
• w




O "*^
h- •»*
rsj U^
O •
• O
w


^ ^>
-— sp
v4 <4~
• •
1 O
^



r*j ^*
«- r— •
v^ v£t
, *
t t'1
s_/


r*. ^\
F«- 1'
rv o"i
* *
I '-1
i^


t" ^v
*M r- •
>T --
O •
• —*
\^




Ul />
i^« ^-
«-, o
O •
t •""
v^




0
(L

y
X
Ul
t/1

-------
4-50
01
01
•w
o
L
o
Id
•-.
£
••4
U
T>
*
U
L.
«*
3
T*

!*
C
•H
••4
C
t.
s:
/•s
Table 4-11 (Continue


(O


flt

03


TJ
CC


55

O
T"
W

•T*

CC
h-
W
»
^^

O-
or
o
cc
»•-
Q.
C£
O
a:
i/»
O
OC
a. /•*
0- L?
«* _j
QC w

ft* O
a o

a. i-
flr 'J
Cj C
a
a
G.
QC
O
a
•
L. •
•n t
a. •
* a
O f
•o
c

1- «.
C* t
• o

-< O
o •
• o

•t CO
o •
• o


f **J

m M
t f*J

«? rj
1 ••

O •
• O

M&RYl'JO -.01'
(0.1,

rf O
o •
o «
o
o
-O x
o •
0 IM
o
o
1
o •
O r*
0
o
fl
1
fl QO
O •
o -«
o
1
rn cc
0 .
O -^
o
1
O t
o «•.
o
t
o •
0 "1
o
0
INCIMCOfl -.000004
(2.«(

h-M o" ,., o i"^ -"- o- wi -4 eo^.
t* O • O* O • O * C5 * •• Ot
O • O • •-• O*4 • O O O * O rw
t |
t
ff» C* in ff» mm inO rjO Ocs h.ff* o^
• t O* O* Of Of Ofl •* O*
O »O t *+ O»* «O *3O f*1 0**^
• 1 1
1
ifl C t O* Ofl Ofl Of •• Ofl
o • o •»•« o«-> t o oo *r om
I
ODC* CDCO ccr- m* ^-0 oo ^c- o>^
• f ff O* O* •• O* •• C •
f O O •fMO^* •*• • O )*^ » *^>
1
^ en cocc com i-j>r inr* ^ M ^-o >ra:
»• •• O* Of f« O* •• O*
(O O f—O"-* ** *O <^J • ^
1
*f f« Of CT • ff Of f* Of
i o o f •«-• o •-" ro* fiM (M • - Z 4. x:
i^ ui >*r O — 1 *
>- h- UJ u- a LJ QC £
«**5L3OS'C1X>
1*>U- <
-------
                                   4-51
in  (49),  with the exception  that  the highest  single  hourly ozone reading




during the  two-week  reference period was used  to characterize exposure to



ozone rather  than the  average of the  fourteen daily high  readings.   The




coefficient on ozone in this  equation increased in significance suggesting




that  single ozone "spikes" may play  a role in  precipitating respiratory




illness.   In equation  (79),  the average daily maximum 1-hour reading was




replaced  with the average  hourly concentration  over  the 336  hours.   The




coefficient on  ozone was no  longer  significant.   This  is consistent with




the  belief  that  peak  exposures to  ozone  are more likely to induce acute




illness than constant exposure to lower levels.




     The  health data used in  this  report do not lend themselves well to an



analysis  of possible  lagged  effects  of ozone  and other  air pollutants.




This is because.Individuals were asked in the HIS to report  only the total




number  of  days  of acute morbidity  during  the  two  weeks prior  to their




interview.   The  actual  date  on which  a restricted activity day,  work or




school loss day,  or  bed disability day occurred was not given.  This makes




it  very  difficult to  select  an  appropriate  lag structure to  test  for




delayed effects.



     Nevertheless,  in  equation (80)  an attempt  was  made   to  Identify a




lagged ozone  effect.     In  that logit regression ozone was measured by the




average   daily  1-hour  maximum  ozone  concentration   during  the  two-week




reference  period  prior t£ that  for  which the  individual  provided data on




acute morbidity.   Thus, if  an individual's reference  period  was July 15-29




in  1979,  ozone was  measured  in (80)  by the average daily maximum  at the




monitor nearest  the  individual's  home  during the two-week  period  July 1-

-------
                                   U-52
July 14.  The problems this creates are clear.  A very high reading on July



1, for  example, would influence the average  daily  maximum calculated over




July 1-14.   This  average would in  turn be  used  to  explain acute morbidity




that may have taken place as late as July 29, nearly a full month after the



peak episode.   In  view  of this  problem,  the results  are not surprising.



The  coefficient  on ozone  was  statistically indistinguishable  from zero.



The  signs  of  the  smoking  dummy  and other independent variables  were



comparable  to  those in equation  (52)  which is identical  to  (80)  save for



the ozone measure.  This crude attempt does not suggest that lagged effects




are  unlikely or  unimportant.    Rather,  it  demonstrates  that a  data base



built  on the  HIS  is not  the  best  tool  with which  to  investigate such



effects.




     The  evidence for interactive or synergistic  effects between ozone and



sulfate  (Equation  (81))  and  ozone and  suspended  particulates  (Equation



(82)) was not  significant.  A  number of other possible interactive effects



were  explored  but  are  not  reported  here.   In  none of  the  cases were



interactive  effects   a   significant  determinant  of  respiratory-related



restricted activity days.



     As discussed above, health and other information about individuals not



at  home during the HIS  interview was provided by  their spouse or another



adult present for the interview.  Such information might be more subject to



error than  if  provided directly by the individual.   This may be especially



true for  cause-specific  acute  morbidity where information about particular



disease categories is being sought.  This suggests the analysis of

-------
                                   4-53
 respiratory-related restricted activity days might be sensitive to a split




in the sample between self-respondents and others.




     Several "basic" regressions examining respiratory RADs—numbers (49) -




(51)—were,  reestimated  using  as  observations  only those  individuals  who




answered for themselves  during the HIS.  Equations (83)  - (85) correspond




to  (49)  -  (51)  above,  with the  exception of  a slight change  in the  way




temperature is measured.  The  self-respondents  sample  was  about 30 percent




smaller  than  the  corresponding regression when others were  permitted to




respond  for  absent members.   In  each of the three cases,  restricting the




analysis  to  self-respondents  increased  the  coefficient  on  the  ozone




variable by  about 50 percent  and  added to its  significance as  well.   The



coefficients on  ozone in  (83) -  (85)  are significant  at the  10 percent



level or higher.   The association  between nitrogen dioxide and respiratory




RADs  in  adults   was  less  .significant—compare  (84)   with  (50)—but  the




earlier  finding  about ozone  and these respiratory RADs was strengthened by




the respondents-only analysis.




     A  final  sensitivity analysis concerns potential  biases  introduced by




errors  in  either air monitoring or the  characterization of exposure based




on  the  monitor  nearest  an  individual's  home.    Both  can  result  in




individuals' actual exposures  being  different, and  perhaps  significantly




so,  than  those  assigned them using  the procedure  in this  study.   The




errors-in-variables  problem, well  known in statistics, can lead to biased




and inconsistent parameter estimates in ordinary least squares regressions.




     To  test  the  sensitivity  of  our  results   to  possible  errors  in




characterizing individual exposure, the following Monte Carlo procedure was

-------
                                   4-54
adopted.  First, for one of  the  respiratory RAD models, each individual in




the  sample  had  his or  her  ozone data  randomly  disturbed  by  adding or




subtracting  to  it  a  term drawn from  a  distribution  with mean  zero and




standard  deviation  equal  to  that  of the  "true"  ozone data  for  the  4,800




individuals; negative values were  set equal to zero.  Then a regression of




respiratory  RADs  on the randomly  disturbed  ozone  variable  and  the  other




explanatory  variables  was  run.   This  was  done  fifty  times,  and the




resulting coefficients were  averaged.   (Results are not presented here but




are available from  the authors.)   The mean estimated ozone coefficient was




0.6221 which is substantially smaller than the "true" coefficient of 1.5926




estimated for  the  undisturbed model (this  is the  downward bias  one  would




expect).'3  This underscores the sensitivity of parameter estimates to the




measurement of air pollution and other independent  variables.  It  suggests




that  the   introduction   of  reliable  personal   monitoring  devices  may




significantly  alter estimates  of  the  dose-related  effects  of  ozone and




other air pollutants on the population.



     The  sensitivity  analyses  above  take  up more  space  than  we   would



prefer.   But  it  is  important  to examine  the  robustness  of  the  earlier




finding of a positive and,  generally, statistically significant association




between  average daily maximum  ozone  concentrations  and respiratory related




restricted activity days.   These findings  did  not  change appreciably when




account  was  taken  of  possible seasonal   variations  in.  ozone  and   other



pollutants,  interactive effects,  respondent  status,  confounding  due  to




pollen,  data  completeness  requirements,  informational   cues,  sensitive



populations,  or  other  effects.    Lagged  effects  and distance-to-nearest-

-------
                                   4-55
monitor variants  were also run  although in both cases  the tests were far




from ideal.  Overall,  there  appears  to be a positive and often significant




association  between  average daily maximum 1-hour  concentrations of ozone




and the  incidence among adults  of  acute respiratory illness  resulting in




restrictions in activity.








     4.2.2  Children




     Analysis of  acute  respiratory disease  in children  is also broken into




three categories—restricted activity days, school  loss days (for children




six years of age or older), and bed disability days.  In each case, several




different  combinations of  pollutants are  considered using  both ordinary




least squares and logit techniques.



     Equations  (86)   -  (89)  in  Table  4-12  explore  the  determinants  of




respiratory-related  restricted activity  days in  children.   Although its




sign  was  always  positive,  in neither  combination of  air  pollutants  or




functional forms  was  ozone significantly associated with respiratory RADs.




In  (86)  the  chronic  limitation dummy and the temperature variable were the




only  significant  regressors.   In (87)  nitrogen dioxide was  positive and



significant  at the 10 percent  level,  a finding consistent with the earlier




"all  causes"  analysis.  In addition  age  and  its square were significantly




associated  with  respiratory  RADs  in  (87).    Equations   (88)  and  (89)




replicate  the two previous models using logit regression.  About 7 percent



of  the  children  in the sample  had at least one respiratory RAD  during the




two weeks.   The air pollution variables behaved  the same in the alternative




functional   form,  although  carbon   monoxide  did  enter  negatively  and

-------
4-56
*
to
w
*
y»
•*»
O

^b
u
e
**»
u
•4
a
U1
a*
ce

>.
«*
•«4>
••*•
^
«
Vt
O

^
t
•c
C
C
I/I
ut
a
«a
o
o
J=
a

w
~*
-»•»
**
u
tf

V
**
u
u.
*«
tft
t
O:

t'
•
C
L
i— #
£
U

• t^
i*)
(
*T

a.
j
n
^

^
w%

X ^
a —
IS! i-l
oe o
0 C
*r j
Ct w



a
w^
fy
2
•I
ae





a.
1/1
a:
a
Q
&
K

> t
t.
fl
>

•
&
T3
C
^*
m**N f*<^ tf> S* *4 ^>
^- a — K. — tr -^ r»
f*4 O% *4*O I>J **> (+1 

*r<^ 0"- — < «*^ oo
1 **f *^ w • w

t


o^^ f*j^ (**^ ar^s
sf — ^OO OOD ^^j
tf*-* OCT» t^dD rwi^
* • t » 0 • 0 .
*S^i tf»-* fO O<^
1 w w w * w
t


*»*•% y^^^ «^% r>»^^
—135 mff« «*>r iw^t
•J ^^ n
• • •• o» o*
«>*m rs*o OO OO
t -w w t w • s^
'


•O^* ^^^ r*^^ (^*^s
T^»^ -«U1 O/M O'n
»9h» ^Jf\j — -it f^*«
* • •• Q» Oi
** O O Q o o

\^ C3 v^
*
t


^»^ ^ /^ f>4^% f*-,/-\
«*- >* -O -*rf% r^U^
• • • * O • (-» *
in o o o o o
w *^ • >-• O s^
t •


»—
a
UJ

a. o o o
uj ae cr a
H* «£ Z ^
„ ^ fi
*? *> sf 0.
^ a ^ i^
h» ^ ^X^ ^>^ <^^N -C^1^
*P "^ *• ^ ^ ^ ^ ^
^*^ >r^ *oro csi i/% mrv*)
^ * O « t • • * Ot
OO t»* (S< |»« OG
t w ) vx w w O ^^
0
0
1

1 1 <7» O (O *• i^ O
*^ L-^ «•» IT *o r>»
t i • • O •
*M 1 -^ O O
>w ^-^ O s^
O
o
1
O^ m^* o^> *»^ m*^
ui^ ^ --, »^«» r^wi -dirt
"^^ ^C* *i*^ •*•
1 1 9* •* -o O* O f*
-i •* mo •* O
O « O • O i
• O • «• O (M
w 1 w O w
O
O
1

*• ^ U^*^ r^ *^ o /*, fi» *•,
•f m sr-*r o^ r*m oo
**Jff N*^ v»**^ -^^ •"•C*
O« O« •* •• o«
O^ ••* *\! {^rt O«rt
• W | w w W O S^
o
o
1
1 1 •* s*. ^ **, *~ s\
1 ! ''^ if >J C* ^ T
t • • • O t
c-i 1 (^ O O
\S ^J O %^
o
(?•(
1
(%J>^ i-^-s U1^ ^^ O^^

»^m a;f^ r*j is* »»-T r,rvi
Of O' O. Ot 0.
O-* O^ »o »O O**
Ow .w w w O — •
• ' o
o
•
1
1 1 « ^* o* "\ O «
' 1 ^r i^ m u*\ rvi ^3
^' >O %T ^^ fS4 & • \^ o ^^
0
o
1

° 2
ca o a
— — — U 0
0 0 3 — Z
ee i* u.' r rj
2 Z u x ij
'SJ O « 'u ^
Z LJ tt (^ »
p^^s i^^^ O<"^ *o^\ (Cy^ ^^.
i-J!? 00« *. 0  • •• o. .0 o .
~* ' " "* •* • ~ • s ^
W 1 ^« s_f i i_* ^* ^^
' *^ w 1 ^^ »«^ , w
t

**^\ ^>^ *r/^ tM^s o-*% i^-fc
I^*A ra^a *»e» sr«o r-»^o !T
« rsi r-. ,-n *^l» w » w v^ » -j
1 t


m^ en^. o« c-<^ *. ** _
miM o-^o OCR ~< 0>w I, **
_ m ^./vi i»r- 'vo — 0 _. —
O« Or •» o* es. -"CT-
•2 S2 2 *- ?s oo
. • w • —
i p


IT/~ ei/^ ^>^ ^/^ — ,-, ,__
-a: f-u% * ^ ^.^ ^ „ ^2
s^ s-. *•: o-^ --. J5
re °s -i ,-s c ==>
*^ • Si^

0^ ^-^ .o-. r. » M^ _
JC - -S "Z "' ° ° " 0?
stse:-': £":*•: »«
OO OCS «.m ixJ 0 ra
»v^O**y *^ !w ^/ W-*
1 . ' w
1


O^ _ ^ * ' '-^ ^
>T s»x» H.-" IML.- -».C ~ «i
"*** —f* %T*^ *OO *-» *^r^
o> O« •• o« o« a.
./M o— i- c-r .-> „
1 w . w w . s^ ^ " ~
1 ,
)



5 •- ":
" Z X 1*1 £-,
^ -J « iv "
S ^ ? z =• =
9 T *• » 3

-------
                                                                                     4-57
                            -^ »"•     ry  a$
QC  O
r  c
t*  ^"»     »O ^
f^  •*     CD C*
r"  •«•     r*J *••

O   •     O   •
o  o     o —
O  W      • *-*
                                                                         *•*     V*  ^     ^ f^
                                                                                                     r»  *s     »-< *•»     ^
                                                                      o   •
                                                                                LA  fn     ^« «^     U*>
                                                                       o   •
                                                                       o  o
                                                                       O  w
                                                             IT. o    •*•*

                                                             o   •    o   •
                                                              • o    oo
                                                              t  %-•      » *^
                                 ir.     «o^r     rir-

                                   »     o   •      •   •
                                 m       .tr-     *-•-
c.  *-
t* M
rr  vT
                            a*  /^     u^ '^     ff*
                                 »     O   •

                                f*     O O
                                                                                                                w- **    O
                                                                        o   •
                                                                        o o
                                                                        o ^
                                                                        o
                                                                        o
                                                                         t
                                                                         I
                                                                        o   •
                                                                        o  o
                                                                        o  <->
    c.
    w»
    ft.
    O
                    **• o
                    o   •
      •     o   •
     o     o o
     >^     O w
                                         o   •
                                         O —
                              o
                              r*>
                              o
                                         o   •
                                         o  o
                    o*
                    o*^
                    O  w^
                                                                                                                                           o >r
                                                                                                                                           «T -»
                                                                                                                                           O   >
                                                                                                                               /^     ^ /%     f«-  ^^     O ^^
   (Ni     c **"•     •"  ^~
   O     »• ir>     r g  r-
    .     O   .     O   •
   (V.     O O     O  O
   \J      t <*S     O  W
           I           •
                                                                                 o    •     o   .
                                                                                 o  o     o  o
                                                                                                                 oo
                                                                                                                  • s-/
                                                                                                                  1
         •o
          c
2
-1
D
                                        O
                                        err
                                        2
car
z
a
                               O
                               Qt
o
u
X.
t-
u
                                                                                      3
                                                                                      n
                                                                                      £.
                                                                                      _i
                                                                                      CL
                                                                                      r
                                                                                                                                                 a
                                                                                                                                                 X

-------
                                   4-58
significantly  (10 percent  level)  in (89), for which  there is no plausible




explanation.




      Equations  (90) -  (93)  explore school  loss  in children, the first two




using ordinary least squares  and the  second  two using  logit  techniques.




While ozone was  positively related  to  respiratory school  loss when entered




in combination with  sulfates and  total suspended  particulates,  it  was not




significant in  (91)  which includes  all five air  pollutants.   The chronic




limitation  variable  and Income  were  significantly  related  to  children's




school  loss,  the  latter   suggesting that  respiratory BADs  are inversely




related to annual household income.  Ozone was positively and significantly




related to respiratory school loss in both logit runs.  In those equations,



the  race  dummy  became  significant  and  indicates   that   black  children




experienced less respiratory school loss days than white children.




     Equations  (94)   -  (97)  complete  the  analysis   of   children's  acute




respiratory  illness.   They  deal  with  bed  disability  days,   using  both




ordinary  least  squares and logit.    In  (94)  ozone  was  positively  and




significantly  related  to  bed  disability days when entered  In combination




with  sulfates  and particulates.   Although  its  significance  was diminished




when  other pollutants  were  introduced in  (95),  the   coefficient  on ozone



remained  significant  at   the  10  percent  level.    In  the case  of  bed




disability, the  logit  analysis paints a slightly different picture.   In




(96)  and  (97)  the  coefficient  on  ozone was  not significant  at  the  10




percent level.  Thus, ozone may play a  role in explaining the number of bed




disability days a child experiences, but be a less Important determinant of




whether  a  child will have  one at  all.   In the  logistic  specification of

-------
                                   4-59
children's respiratory bed disability, nitrogen dioxide entered positively




and significantly at  the  10 percent level.   It  was not significant in the




OLS specification.




     Using restricted activity days,  school  loss  days,  or  bed disability




days, the  analysis  of acute respiratory disease  in children was dominated




by  the  chronic  limitation variable  and  the average  maximum temperature.




The latter suggests that  the  higher  the average  maximum temperature, the




fewer the number (or the less likely the occurrence) of days of respiratory




disease.   The importance  of  the chronic limitation dummy is  clear.   The




effect of  maximum  temperature  is less so.   One  reason f.or its significant




negative effect may have to do with colds  or respiratory infections.  Were




these illnesses  more common in colder northeastern and  midwestern areas,




then  average maximum temperature  would take  the  sign  it  has  in these




regressions.   This  suggests the desirability of  additional  work along the




lines presented here, but  in which  more  detailed analysis can be conducted




of possible epidemic effects.








4.3  Chronic Respiratory Disease



     For   several  reasons,  chronic  respiratory   disease   (CRD)  and  its




possible  relationship to  ozone and  other  air   pollutants  is  of   special




interest  in  this  study.    First,  unlike the analysis of  acute morbidity




(where clinical  studies  can provide valuable information), epidemiological



studies  are   the only  direct   source of  information  about  the  possible




deleterious  long-term effects  of   exposures  to  ozone on  humans.   Also,




chronic  illness  is  in  several ways  more  serious  than acute.   Even  some

-------
                                   4-60
serious  acute  illnesses  appear  to  be  reversible.    However,  CBD  by



definition  implies  a  continuing  state  of  ill  health  which  might  be



moderated but not easily "cured."



     Furthermore,  chronic  illness may .predispose  one to  acute morbidity.



In the analysis of acute morbidity above, the dummy variable indicating the



presence  of  chronic  illness  was  always  positively  and  significantly



associated with acute illness.  Thus, even if ozone were not a direct cause




of  illness  in  healthy individuals,  it  might still   have  an  important



indirect effect if it increased the likelihood of chronic illness.



     Finally,  unlike  the  case  of  ozone  and   acute illness,  there  are




virtually  no conclusive studies  linking long-term  exposures to  ozone  to



chronic  health  impairments. ^a    This  has  been  ascribed  in  part  to



inadequate  data and methodology.   It  is . important  to  see,  therefore,



whether  the larger  sample size,  the  matching  of  individuals  to nearest



monitors, the  creation  of  longer  term  pollution exposure profiles, and the



inclusion  of residential mobility  data—all  innovations  in  this report—



alters the assessment of ozone and its relation to chronic morbidity.



     In  the following  analysis  the  focus  is  no longer  on a particular



two-week reference period, but  rather  on whether or  not an individual has



CRD  that  is correlated with long-term exposure  to ozone  and  other  air



pollutants.    Accordingly,  air   pollution  measures  are  specific not  to



two-week periods but rather to longer periods of time.  In the first set of



models,  annual average  pollution data for  1979 were  used;  subsequently,



analyses were  conducted using multi-year average concentrations that may

-------
                                   4-61
begin to  characterize individuals'  long-term exposure to  ozone  and other




air pollutants.








     4.3.1  Adults




     With the  exception of  the  period over  which air pollution data are




summarized, the basic model used to explore the determinants CRD is similar




to   that   used  above.      It   includes   socioeconomic  and   personal




characteristics, meteorological information, and certain other variables as




well.   Ozone,  sulfur dioxide, carbon monoxide, and  nitrogen  dioxide were




summarized  by  their  1979  hourly  concentrations.    Sulfates   and  total




suspended  particulate  matter were  summarized using  1979  twenty-four hour



readings.  The  dummy variable  indicating  the  presence of  a chronic illness




was excluded from the  list of regressors,  since  the  respiratory component




thereof is what we are trying to explain.




     Equations  (1) and  (2) in  Table 4-13  present  the  first analyses of the




determinants  of  CRD.    Ozone was  measured  by the  average  daily  1-hour




maximum reading over the year  1979,  as were  carbon  monoxide  and nitrogen




dioxide.    Sulfates  and  total  suspended  particulates   were  measured  by




average daily  (24-hour)  concentrations.   The sample  used  in  (1) included



7,210 adults,  537 (7.5 percent) of whom reported having  CRD.   The sample




used  in  (2)—which  is  smaller  because  more air  pollution   data  were




required—included 5,891  individuals,  of  whom 435 (7.4  percent)  had CRD.




In  neither equation  did  any of  the air  pollutants  exert a  positive and




significant  effect  on  the  likelihood  of  CRD.   Income  was  negative and




highly significant,  while education was positively and significantly

-------
4-62



















£
ID

C
0
t~*
0
a

CT*
r«
a

••
">
_4
3
*•*
"•
O
C
o
e
<

Or
t/1
It,
a-
^

(.
O
+*
ft!
{-
>**
a
w»
ct

u
••4
C
o
u
£
U

«
en
i


n
— <
^
(V
i—



/-%
T1
W






^%
pu
W




/•N
*o
w



l/\
w






xf
^






/"\
f*l
Xrf





^
r>i
^/







^N
•-«
W










•M
0)
•1
0

0
t/» ^
OC P-
r •-•
LJ L"
V> Q
OC ^
0. w

a
i^ ^
QC P»
r -*
u o
f» a
QC -J
a •-'

£ ^

*/* o
« -1
CL W

a: p-
w o
trt O
at J
C w

0-
u} •*•»
•v- »-
(-•  a
cr _J
a. w


a.
c» <^
QC t—
r •-
W LS
i/) C7
QC ^
a ^

a.
Of ^
I •"
0 O
LO O
IV _!
a. v


a.
W1 <^
OC h-
X -•
u o
oo o
u -J
/S lfcy


fl
•0 i-
>
•
a t
a- a
u o>
•o
c
k— «
*^ ^ t t ^ ^ t 1 U"' ^ ^ ^ w* ^ *O ^X
«^ ^ ft v> f™> i i »o >?• r*» *^ ro ^o u^ o
^ ^ ^ ^^ ** *£ ** CT ^ »O ** <**
• • •• 0*0*0*0*
f"4 |t || *o/^ *o^ CT»/** O'^
u-^cr? OU* ir iw *-r»j ****
*••• o»^»ft*o»
rv^M O OO»O«**OO
I W *-/ * v-* s^l^H/•^^




O<«* r. ^ | 1 || **/^ e*>^ fl tt
*^ ^ (T* >O It t 1 ^ir'  • W • W



tr, ^* «*** ii if v*rg tric^ -^o — i cr>
O ** •*" "-5 ^*iO^ct<>*l^Or*'O
• • ** O»O*O*O»
f*-, **J O O O • O • <«* O O
1 >wy ^> • v> ««ii | ^/ * v^




o»^^o^ it it sr-^>r^ it it
^ u% ^t r* i i i i f) CD «£ u^i ii |i
f^OT*^^ ^^u.iO<*>
*«>« o • o «
mo o o o o o
f W W e w • w





c*.
OJ
o^^m^f ^s T« «^ ^*
orooo oooo
UJ<<4«44<
>-Z2ZZZ^^
2(^»mm (L«ro 'W
i-*ciaai^i/ioz

^4
>^ «O *O •-> N*= u^ ^f^ O9*
*« ^ ^M *^> tr> 1/1 ^j CD sn c^
«• . .- o • o* ••
O 1 «-» • O O sT 1 O
*^ ^r t ^^ O W N— '
o
•
1
P* M <^ ^ ^ >r P« r-, _ e
* * * t O* O* •«
o i «-i * o o ir i1 •*
0
«
1
^» C** r^ W? 1**' tPi ^J ff W" Q
.• •• o* o» ..
O 1 «-" « O O s? 1 **
w W 1 w C: x^ >ta^
1
1


•^•/^ rg/^ VO^N o>^ ^J»^N
— o <^ r* sr -^ -" 1-1 -« »•
• • • * o« o» •»
o i <^ • o o in 1—4
V_/ ^^ | ^^ ^ ^ ^^
o
I


!••*><-» vt>^ m *^ o *^ ^. '^t
mm "Ofj -^«*) r*tr- »-«o
>p^ ff* •« *^ U^ I/* **! T> ••'"1 Cl
* • e* O* O • *c
O 1 — * • 0 ^ -J- \ -*
W ^ | W O w ^
0
1

f^t <0 O — " mu? *NJ »* «,n
w; r^ t-« C" V>^> o*"* ep^> cn/^
>*1> •sf'C f^-lA P*=*O O11^
p v^1 — •
• C *t O « O * e e
O 1 »-• * O O O- 1 *-
w w I w O w w
O
•
1


0 Z
33 O O O
-* u. Z u
~2 ** ^4 ^
LU 31 >•> G
LJ x a u p-
< IU  Z *H a.

a. r*>
«) *
* *
^^
s^



•0 -^
W *C'
** m
• •
_,
x_/



!- at
e t
—



•>4 rn
•-«
Ku«




«t x>
C" C
» .
«•*
\^<




t*' <•«>
m r1*
•^ r*^
• •
i— »
v^



rr —
— • u-
« •
*«
\.x




^ ^^
•0 P^
*~> f
e e
,_>
\^f








•c?
i/l
k—
tf
0-

-------
4-63








i^
**
'fl
Q
C
o
=
o
a.

(^
^»
>>
"1
3
* ^~
i ^
<-i a
•J^ _J
iX -^
0.

a. •«*
iXt H-
C£ *-<
z o
(_j •— I
o-> -4
(TC Virf
O.
ui h-
ot "-«
r o
0 Q
v> w
CC >H^
a.

*x> H.
(X ^-«
Z 'J
•J Q
1/1 j
•r w
a,
•
u t
> ^
«
u, t
it a
o w
•o
c
*H
II •* ^1
II O «T
T1 O
O •

O
t
II 0 ^
It 0 —
f* A)
o •
o m


II 0 «
•J* O
o •
O wi
O w
o
II O /^
»4 OP
o •
O f*>
O w
•


II 0 ^
II O •*
7* 0
O <*\
O \^
o
•

II 0 ^.
II 00
f* .-e
o •
O /*!
O w
*

O -f ?** -^
rn %yj^ 'Ji'O too
o* o « o* o • o * o*
O *"4 * W O O O O O O *O u^
1

<0 * 0^ vO ^ ^^ « /-» -00
-^^* r*m <>«o <^f* ad ^-^4
fsj vQ ^ O in ^ ni ^ "* 'NJ • 04
o* o* o* o* o* -r *
OO *ro OO OO t*4 -ON>
1

m« ^^ ** *~ co^ tn^
^ rvi in** rMrsj *f tx \P-O • T3
Ot Ot Ot Ot Ot O*
o -^ «<*i oo oo oo *OIA
• N^ ^/ • W • S> • **^
1

OJ *O -* O ^f *O fO O3 J1 ^* t M
oo • rn oo oo o^ *<)^
1 ,
•

^7»^ M3^ jnv-% 03 * co^ >0^
ajn^ v^3! \r *f fM->O rjr^ p*r»
*r M i.\ -4 <-^ (^ ^. OD «% -^ » co
o— •<*> oo oo oo ^m
• V> N-* *-* ^o--» oo
i/N JD <*• N|- -rr* *>4r*l rvj -^ 0^4
•%j ,<•> ,f r^} ^ ,fn fry -r5 CT* v^ * fM
Ot Ot O* Ot O* •* »
O O • f*^ Q O f^ O O •* *O ^»
i s^ s> *\^ • s^ «s^
1

O • O • O « O • O t OJ •
t\> N-* «>^ * *-> »V^
1


£>o m— rt-i^ ^4-n ^>o o2
^•r -./•^» .^ .« cna? rj>»« *^i
O I O* Ot O « Ot J ^O OO O*1* -Of^
• ^ N^ «^rf t^y • S^
1


r
£ a 3 IN
-« u
M o £ uj £ a
u a iu oe 3
2 -u >- a. c u. z

-------
                                   4-64
related to  the  presence of CRD.   This latter effect  was  unexpected since



more educated individuals,  by being better  informed  about health matters,



might be  expected  to have less illness.   It may be that these individuals



are more  likely to  have  had  CRD diagnosed  as  such,  than  those  with less



education and, perhaps, less access to medical attention.



     One puzzling aspect of equations (1) and (2) is the performance of the




age measures.   One  would expect  older  individuals to have more  CRD.   One



explanation  for the  insignificant  coefficients  might  be  the  presence of



collinearity between age and its square.  Diagnostic tests for collinearity



revealed this to be a problem.  The regressions were rerun using the square



of age as a single measure of its effect (because a non-linear relationship



was suspected).  These results are presented in  (3) and  (4).  The square of



age joined education and income as significant correlates of CRD.



     Two additional measures of ozone exposure were tested.  In (5) and (6)



ozone was measured by its annual average hourly concentration in  1979.  The



results  were similar  to  (3)  and   (4).    In equations  (7)  and  (8)  the



regressions were rerun using the single highest hourly reading for the year



as  the  measure  of exposure to  ozone.   As in the  analysis  based  on annual



average daily maximum  ozone readings, income, education,  and  age were the



only  significant  determinants  of  CRD.    The  fact  that  the   number  of




cigarettes  smoked  per  day  was  not  significant  is  unsettling.    Later



analysis sheds  some  light on this matter.



     It should  not  be  surprising that no significant relationship appeared



to exist between CRD and ozone and other air pollutants  as characterized by



1979 values.   One year's exposure to even high  pollution levels might not

-------
                                   4-65
be sufficient to  cause  or contribute to  the  onset  of CRD.  Annual average




values  for  a single  year,  1979  in  equations  (1)-(8),  are used  in some




studies  with  the  implicit   understanding  that  they  are  proxies  for




multi-year exposures.   If the  values for a  particular  year are atypical,




spurious correlations can arise.




     To minimize  the  likelihood of this occurrence,  an  air pollution data




set was created consisting of  annual  average  pollution data for as many as




six years,  1974-1979.   These  data were assembled  on a monitor-by-monitor




basis and matched to individuals, thus enabling us to assign to individuals




longer-term  average  air  pollution concentrations  than  have been  used in




previous studies.  This  sheds  light on the bias that may result from using



1979 data alone to explain CRD.



    ,Equations (9) and  (10)  (in Table  4-14)  present  the first analysis of




CRD in adults using the multi-year averaged air pollution data.  The sample




in (9) consisted of 7,240 adults including 555 reporting CRD.  Six thousand




and five adults  form  the sample in  (10),  468 of  whom suffer from CHD.  In




the  two  equations   all  pollutants  but  one   were  measured  by  their




(multi-year) arithmetic  mean.   Total  suspended  particulates were measured




by  (multi-year)   geometric mean   concentrations.    Because  no  data  were




readily available for sulfates for the years  1974-1978, sulfur dioxide data




were used instead.



     The introduction of air pollution data averaged  over longer periods of



time made a  considerable difference in the results.  In equation (9) ozone




was  significantly associated  with CRD  at the  10  percent  level.   Sulfur




dioxide also entered positively and was significant at the 5 percent level,

-------
4-66

>J
«l
•H

-
ifl
C
•u
VI

••5
C
•3
'0
**
"1
O
c
o
••4
«e
3
^
a
"X
t.
r
X
1
•»
^
£
l/»

.^
"3
')

C
O
e
*t

^
o
w
•1
4
•x
w.
o
**
N
u
oi
'1
'<»
Of

'J

c -*
o >•>
V. V
.£ »-*
tj T
C

o
o4






,-*»
WV
r*
W





*^
*r
r-*
W




y-S
m
—



(*>
t j

w




.-%

_^
^_y





^
0

SMI







y-H
J*
W






•t

M«
*
•o
o

a.
ol ^>
« H-
E M
0 O
 ("^
ac >-

w • j
e* -j
a \^


a.
IX) «-V
cs *—
X. **
CJ '41
« ™J
Q. w


Q.
\/l *"%
CC *—
X *-*
W 'J>
ac -j

Ut *"*
on >—
31 *-*
' J L?
ivi C3
Q_ s^


a.
0£ >—
T >-*

"i C!
^ J
.T. w

a.
^/i "•»
~« *—
X -*
0 -J
^) U
f J. J
JX >^


a.
•,^ **\
QC >—
Z ^«
O ij
v^ O
^ J
Q. W

1. •
tl L.
>
•
a •
i* a
a 4*

c


r ffl *M -o
>~sj t • * O •
1 r*» r\t rg O O




(M ^ >O ^ II tn ^
^ 04 in 1*1 | 1 04 O
•fl « tA *M 04 O
• • f f O t
^f M f*i o« a a
t
l


O >"* o* /•> || <^ x^
04 ^c -o ao ii *x r»
(^ f- f» r«j h- ^
• . « • o •
in rM m O O **




m n r- ^ II f* (T*
p*. Q %Q *O ->f 7*
e • * « O *
xf <*l ^> O O O


rj ^ II m ^ r>- /-.
u"\ rn It c»f m r^ LA
o« O* rj r» ro o
• • • • 0 •
>r '^ o* o o oi




c^ o* ii o •** ^n 10
^ • if-, *^ f*i ^ i-^
• * c , O •
»r m o4 o« o o
l <*y i s^» • s->



<^ /^ >^ /^ || lA /^
«rj vO \T< ir\ II 11 ^
rsi i^ in ^ <^j ^
. . • . 0 9
-* **i r- o o o
1 W Vw* • W




H-l ,-S 0-4 ^% || 04 ^
^J CO >^ (*> |1 O 00
o4 -f ^ -4"> o4 rsj
• • • * o •
o- m (vi o* o o
I w ^ w « w




Q. > <^ >
MJ *T X <4
(J OC 4 kU
« -< S 13
UJ Z Z Z
o» < «< *t
S "^ ^ fl-
fc^ O Q cfl
O ^ II II II
*1 0 II || ||
fl *^
o •
0 0




O --» It I) II
O <"J It 11 II
04 X
0 •
• 0*




^4 ^ ^ ^ CD ^ | t
a ao -o o o* o ii
04 *^ «i ao m f^i
o • o • o •
t o4 .0 a o
i



C^ ^ (1 (1 i (
<"vj f^ 1 1 1 | (1
ro ^
a *
o -«


^ /•% ^ ^ m ^s f i
h- O Tt '3 '^4 O |t
<7> ui f* ^* I'M O
a • o • o •
0 y* i<% f» o4 r»j >o
O • O • O •
O fM • *-• O O
t w | w • w
1



^ ** 11 II t 1
m ^ 1 1 It II
*rt O
o •
O f"!
t \^




> > >
4 -t «t
cc oe oe
*t «< *<
Z Z Z 
>r^0 *o^^ m^> MO
,-n *•
1 O
1


o os in ifi in f*> •* <3

. « «9 Q t a*
«-* | 04 « 0 *3 -*
"•^ Srf I w O w
^
I

inx-» F"*'-* ^'^ •?-j-*
in r> /^ J f^>
• • •• 0» 0«
O | o4 * O O >*
•w w 1 w O w
O
0
1

o ^ in ^ ^ ^ tn '^
O 00 W* >O (** O ^ ^
• a
 r*j
r>| **•»
• •
1 O


F- ^
*3 -^
04 04
9 •
1 0




<* 
-------
4-67




•o
c
m

n
o

c
o
—
3
•^
••»
o
Q.

t,
in
^
>.
1
1*4
•*


• •
W
*-»
w
3
T*
CI
a.
c
o
e
«
«
m
fl-
irt
s

>>
U
o
*•*
TJ
L.
'£
tfi
f^
Qi tfl
•r4
u f
•*4 X
C -*
O (*1
t. C
-c 
3 ••*
C *

+* ui
c c
o *
U I/I

•^
1
.*•

a-
— «

(3
^->


/•N
vO
**
\_/





^
to
-*







^>
^
v*
W




^-n
m
^"«
^t






/•%
r«j
C





^
,_ i
^
s^






/••»
O

w






r1*
a*








9p

^4
^
•o
0
Z
o.
Wl fN
or *-
X *•*
U |J
l^ Q
cr -J
CL w

a.
»/i i^
oe ^->
r *-*
^j O
« -j
a. w



a
Ul ^N
o* >-
r ^
tJ C1
^ o
£^ ^
a. w
fc
»/^ /^
cr i—
37 v^
u c
to o
C£ _J
O. ^


a
-*

a
on ^
CC h-
— ^*
w o
t/» a
ac _j
a. w

,
u •
It L,
> tJ
>

a •
a» a
a a>
T»
c
*^

£7»o t f ir\a* CD*T )| or- tn o mr» | i | *rro
M^ IA »fl t^ CO O* ** U> v4 r*^ *^ U"1 •^ ^ " ^^
o • 0*0* 0*0*0*0* o> *
• o o ^ o o • o t ^4 o o * ^4 ^^ •*
w O^*^ s^ | *>^ * ^/ ^/
0 1
•

s<3 ^ u^ /•> || P" i<^ it ao'^ *o *"* r\i *-% r- /•% m^N i i i i >y O (J*fN* *#<> tftn f*>^ |l |t PJ f*l
*^<*vJ^CO 00 O <^t/\r*-iOrv*'*^*(7*U^^ • O it »H^- cu v\ ii p^**> a»«* p»O .i*m it ii 11 <^«-
r* •* >** O T* **" f"0 ^ CP W^ ^ ^ w4 ^ • ^
o« 0*0* o»o»o*o* •& »
• o 0*^0^ **-ooo*-»w sj»».
v^ O^>*\^ wtw'^ x^
O I 1
*
* ~ II u>^ «c^ ii »o^ m ^ 0^1 m^ tl 11 tl ^e-
or*. ii ff-f— en*f> (i */ih" oift **j«* ai»»' ft 11 i i f^o
*- r— tn UA m o "tr— ^co^tf'Ocn • ^^
• • O • O » Q«C>»O»O* t" •
O O-H.O^^ »-*»OOOOO Ui-r
w O v_/ t w w y^ » >*^ • i^*
• i 1




vO ^ II ^) <*^ ^ ^ t 1 ^ ^ sJ ^ f** ^ '"J •^ It It t 1 'J1 i. .
C" ^i ( ( r* O *\j *vt It M~ ^ vT «O C**"^ 'M*^ II (t |l f* o
*> ITS r* r- ^ rj r« u^ ^ fi C« ff> fV ^ ^ C^
tO Or^OO »*^oOO«-*«fM -*.--o
O • t
t



**«o ii CD a? cc c* 11 a>\fi o -* ao TI aj o It It l> *-• **
••M^* *•» C^ ** ^ ^JCPp^*^ t ^' <•* O * ^'
t* 0*0* 0*0*0*0* 1-» »
O OfVOO f«*»»NO«^f«»j ^C'*-
\^ O^*^^ v-* ^ • w s^
O 1
*


-O'^ It f^.^^ «C/^ it -C^ P*''^ C^*^ tA/'N (I II (I r-iT"
f»P\j 1 ) f*rsj ^-%» II f»-O ^»*NI OJfM ^^r*-* || II || — r;
«Otrt f* r— (7* <^ r>j^rjr^h-r—r^o * O
O» O • O * O»O»O»Oi v»«
• O O ^J O O * *•* O O O »•* • fJ w s£j
^ O W O \^ **^ • W ' w v-*
0 • 1
*

•-« y^N t 1 f* <*^ ^* ^*> 1 t ^ *™^ *f' '^ ^* **^ "^ ^ I' ' ' ' f V) O
**4 >f || OD *^ OD «"4 1 1 C^ ^1 *N N> •>*" C? »O >^ t | f 1 t 1 r^i **
»^c^ h» o •-• ur fMc7^^r^rJ»fl*i<^* «4 »
o 0^*00 •••oooc?*-* «o^«
^> Osy*^/ x^»v^*v^ w
0 1
•



Z 0 2 C Z
>. z a a «t *M
O^OZ2Zi-»22X
C? H'co>-3;^'^^Q.'iJ<_J
t/1 i/lO^On.OOto«JC3U
>— UJUJ*— •^3LJ3ri*J2T«*Ja O
4O«JO£QuJQC3ODUJ
u_<««r;(^iat>— Q-X^c-a-u-Z

-------
4-68
c
m
«!
«•>
0
c
o
••4
3
^4
O
o.
1.
n
X
i
•«•*
•3
Z
„
if
*•
3
4

tr.
c
o
F
^

%

is
«•
i/i
a
X
L
0
4^
fi*
'-
a
«•
"
U trt
•»4 -r4
C vi
p ^
(_ w*

C -4
•H V
+* **4
C in
o c
O 4»
w t/t


>f
_t
1
v*1

*
M4
.D
fO
^»
















/•N
ru
W






x*.
O

w





n^
f*
•4
^




ffl
•->
*w^





^
h*>
^
s**









i*

— «
a»
•o
0
a:














a.

oe H-
LJ u:
at ^
f^ ^



a.
t/> '•>
cr ^—
I *-i

irt o
or ^j
&• ^

a.
LI *-*
QC fc—
r -a
w y?
t^ C3
O x^

CL
CJC *-
Z M
U l^
io a
c^ -J
a. w


a.
QC ^>
J. •"
u> O
t>-t a
OC -J
& v^


ff
L> «
m c
> T
>

a «
a> a
a «
•o
c
1-4














*>/•* tf>^ )| ^^* «Ni^ II |l (I «. /-. *^» ^/-» r^^, m^»
w*. w% O^> || *T^ f»«K> (t (I ) 1 f*- -*v> c- I*"- -* <• »»•»«• »f <-»
lAODOrst ^4*flC"*^ ^^CE«^f*jnOf1ldj(TO
v^OCL>o ^( t-i »^ fn <"v o «-*»-« r^w^Meiwaw
• • • * o « o • «»•«•• *•* • • •
«•< O 1 O • f^J • ^ ^' 1 ^^ ^^ O f\{ »— i rvj
I s_/ s^ i ^^ *^ \,/«M/v^rrv-^i\^
O
1

••)/% CT*»^ (I sC'^ ^^^ (f ff (1 «*^fc f*>^ ^s^^^ O *^ (%)^»
p*m O «Ni 1) C^«* m*O II It II (Nl'** ••J'O CC ** '*•*!• r\iui
O •* -^ ^ •»• *M f* »* fvf ir ••* i" r«- c" "j -* •_'• *-"
**•• O*O* 0B««O»O*tC
«r^^^ CTOO^ ^*i^-.o<^^io
« o
1

CT»<* mr^ i! *r»n 3*0* i i i i .*y> or>j i^»c ^ir -j-c- p»o
•O **j -*f f* C^ •/ ^^ V^ I V> O W W/
• o
•
1

3^£ ^Iln it ^TO in In ^o* f^o i 1 £ o ^-(^ ^^ *^o o^r
OJff> CT»O lAf>J ^»«O "*».f ^^ •* O t-^'*' r* r+ fM*M n M
«• •• O t O • • • O • O* e * O • O » • •
^O ** »•< •* O *^ O P* t "^ O O • C3 J O * O ***» /**! O
| Sri' ^4 V«/ «^»^ W • >*/ S.^ W/ ^/ O S^ -W
I C*
•
1



t-=
aB>rw>>>>' o s
Hi •* X «T * •! *C r*l O O Ct
o QC ^ w QC a ot *-» it 2.! o
ne^fEOtf*^ rtw^^^'ic
kuZZ2TZ£Z'MUJic>>O
h*44«4^4t.^OXr^iH>^
2r-tfTa,PjQ
-------
4-69
c
'•fl
w
c
o
3
O
CL
u
'9
X
t
••<
T

.,
j>
*>
— t
•3
c
-T
C
,•?
t;
"*
i.
L>
-^
*j
T
^
X
u

**
m
h
u.
,1)
~£
jj
1) .4
•* .)
C X
o —»
t. '•»
-C C
t i 
3 -4
C **

** in
C C
o v
w '"
^.
f^
I



^4
a
TJ
^-













*-*
_*
x-\J
^





o
r- '
w





x-i
•^»
^^
^





^>
ji
«i
XjH





^
r*-
.^
^^







^

•4
V
T)
0
31











CL
l/» «^
u- 1—
^ ***
»-J O
t/i a
OC -J


o_
v^ s*
I i—
, j • «
VI G
^- -J


a.
i/i ^*
Of t—
E *-<
i_j ,_;
"."' n
(X. v-/


a.
S/l />
Ol V—
£ 1— 1
'-) O
VI Ci
flC J
CL V>


a.
C£ ^
C •-"
W O
u"> O
QC -J
a, x>

4
ti u
> fl

,
a •
•D a
ri >
•n
c
*^











co^^ tl ^3^^ 03*** ii *A^* o /•* -o^^ ir>^% { i | | i i *o^
-^f*^ tt O^^ f**^» it O^* C^i4
s^ O s^ • >^/ w w 1 w * w
0 1
*


aa/^ ii in/-* ^/^ || CM /•*> F-I^ jj/-* *n ^ it P^-^* II c*^
srci i*icr»m2 fs*a».-n«(v^-j»o i^ou tcr»
• t O « O • O»Otr3»S"5« ^3» ^«
eg O O a O tO« • ^^ ^ ^ 1 w O ^ v^
0 1 «
*

.^^^ *-«'^ ii ii fk«/^ hv«^ w> ,"* c** '^ T? ^^ ii ii ^n p-^ r^r
mtA 'M'* II I) Of^ C^flD OlA <-^ O *Of> (1 It ft f^ »AO
•^-*r*l'*\J •*> r*> «J *" vf v^ rv| sf> »4 T3 U^ p» • -f
* t o » 0*0* a* 0*0* « « IA »
- "-V r^, ^ \o *^ mx> ii ii (i m o
•**^V \ t rf* f*\ TO^« It ^-O ^f-4 <^i^ C^O It tl t} -O>X
^^ (7* r» O »-s m 'M^PK|r tt i i |i .-••*/
^*O p»«^jff>'B »Htn^*f»*miA*^'O • •*
o* o • o * o » o • o * o • r*- •
OO O*^JOO eOtQo-^««^ ^»*
• V^ O ^^ • 'M' V^|\^l\^ >-<
O I
•



2 o z as
*• 2 O « & *4 **J
Q~4O22<»«ZZX
'""1 ^ >^ ^ £ 4 *J O CL 1J O
•yi LrOO^Oii_>*^<^JOac
^'uiui-iaozujirtjaa
^ (j ^j ^j 2H Q i^ Qg ^j ij Q ^j
'X 
-------
                                   4-70
while  particulates  were not  significant.   Income  and age  (squared)  were




still  significantly  correlated with  CHD,  as was  education although  at a



lower level of  significance  than in  (1)-(8).   In  addition, average annual



humidity became significant at the 10 percent level.



     The  coefficient  on  ozone  in   equation   (9)  was  sensitive  to  the



inclusion of additional air pollution variables, as (10) illustrates.  When



carbon monoxide and nitrogen dioxide were added to the estimating equation,



the  ozone  coefficient  was   reduced  in   size  and  significance.    The



coefficient on  sulfur  dioxide increased as a result  of this addition, and




its  significance  level  also  increased.    The  sex  dummy  also  became




significant at  the  10  percent level in  (10),  and indicates  that  men are



less likely than  women to have CRD.  In  equations  (11) and (12), exposure



to  ozone  was  characterized  by  the  average of the  second-highest annual



readings for the  years  for which data are available at the monitor nearest



each individual's home.  The  other  pollutants  were still measured by their



multi-year annual averages.   The coefficient on ozone turned negative when



this  change was  made  although  It  was  not  significant.   Since  chronic



morbidity  Is   less   likely  than  acute  illness  to be  related  to  extreme



values, the performance of 03ANMAX2 is not particularly surprising.



     We  believe the use  of  air  pollution  data averaged over  a number of




years  to  characterize  long-term  exposure  is a  significant  improvement on



the use of one year's data.  However, analyzing the effect of air pollution



on CRD would still  be  a difficult task even if one had accurate historical



air  pollution  data  from  all locations.   This   is  because  individuals



generally do not  live in one  area their  whole  lives.  Lacking information

-------
                                   4-71
on residential histories., one might falsely ascribe a high incidence of CRD



in a  particular area  (say St. Petersburg,  Florida) to  air  quality there



even though many individuals now living  there  may have lived most of their



lives in  industrial  cities in the  northeast  or midwest.  This  would be a



textbook case of spurious correlation.



     We made use of  the  residential mobility  supplement to the  1979 HIS in




an attempt to circumvent this problem.  Using the supplement, we identified



a group of adults who had been living at their present address for at least



five years  at  the time  of  the 1979 HIS, the  approximate period for which



air pollution data were  available.   A number  of  individuals  in this group



had been at their present address their entire lives.  In all we identified



4,108 such  individuals of whom 309  (or 7.5 percent) reported having a CRD.



The  results  of  some  of  our  analysis  using  the  subsample  of  these




"residentially stable" individuals  is presented in equations  (13) and (14).



     Comparing  (9) with (13), it  is clear that  confining the analysis to



relatively long-time residents of the SMSAs in our sample has the effect of



reducing both the size and the significance of the ozone  variable.  Whereas



in  (9)  ozone was significant  at  the 10 percent  level,  that  was no longer



true in (13).   Sulfur  dioxide remained significantly  associated  with the



likelihood of CRD, although no longer at  the  5 percent level.  Education,



income,  age and  race were  all  significant   at  the  10 percent  level or



higher.  Comparing (10) with (14) results in a similar conclusion.



     It is  important to explore a  number  of  other potentially confounding



influences on the estimated coefficients of ozone.  Indoor air pollution is



one example.  Recall  that  data were collected on the use of natural gas as

-------
                                   4-72
a cooking  fuel  in the thirty-eight  largest SMSAs in the  country in 1979.



After converting  the  data to  percentages  of occupied housing  units using



natural gas, we identified 4,939 individuals for whom we had multi-year air




pollution data as well as data on the probability that the household cooked



with  gas  (all  households  in  the  same  SMSA  were  assigned  the  same



probability).  Equation  (15) introduces  the variable PGASPIPD which is the



probability discussed above.



     If  equation  (15) is  compared to  (9)—neither  are restricted  to  the



residentially stable  subset—one  can see that the introduction  of the gas



stove variable did not reduce  the  size  of  the coefficient on ozone but did




reduce its significance.  The  gas  stove  variable also reduced somewhat the



significance  of   the  coefficient   on sulfur  dioxide although  it remained



significant  at  the  10  percent  level.   Equation  (15)  suggests  that  the



higher the probability of a household's having a gas stove, the less likely



is an individual  from that household to have CRD.   Since  natural gas is a



source  of  both  nitrogen  dioxide and  carbon monoxide,  this  was  not  an



expected result.



     These results  should be interpreted  with  caution,  however.   The  gas



stove data are aggregated  to  the SMSA  level  in contrast to  almost every



other variable in our data set.   Individual  data or data disaggregated to



the census tract  level would be preferred.  As  concerns the effects of gas



stoves on  the ozone coefficient,  two factors  should be considered.  First,



the   sample   size  varies   substantially   between   (9)   and  (15)—7,210



individuals  vs.  4,939 respectively—because  gas stove data  are available



only  for a subset of  SMSAs.   This difference by  itself  could  account  for

-------
                                   4-73
the change In  the  ozone coefficient.   Also,  the  SMSAs for which gas stove




data  are  available are the thirty-eight  largest in  the U.S.   Thus,  the




samples may  be quite different with  respect  to population characteristics




and perhaps  even other  dimensions  which  affect  health  but  which are not



Included among the set of  regressors.   In summary, the  effort  in (15) to




address indoor air pollution should be viewed as preliminary.




     Separate  regressions  were  also  run  on  a  set   of self-respondents




although the imprecise recollections  of others  should be a less serious




problem for chronic than for- acute respiratory disease—it is unlikely that




CRD would escape the attention of a spouse or other related adult answering




for an  absent  household member.  These  are  reported in equations (16) and



(17) and can be  compared  with  equations (9)  and (10) since neither pair of



equations  is  restricted  to the  set  of residentially  stable respondents.




When the sample was confined to self-respondents, the coefficients on ozone




increased  substantially  in magnitude  and significance.   In  (16),  where




ozone was  entered  with both total suspended  particulate  matter as well as




sulfur  dioxide,  ozone  was positively related to  the likelihood of CBD and




was  significant at  greater  than  the  5  percent  level.   Note  that  the



coefficient  on  sulfur  dioxide  was  no   longer  significant  in  either



regression restricted to  self-respondents.




     A number  of sensitivity analyses were conducted to explore synergistic




effects between  air pollutants.   An example is presented in  equation (18),




estimated  over 7,240  adults of whom  555 reported a CBD.   Ozone and sulfur




dioxide  were   both  entered   individually   along  with  total  suspended




particulates  and  an  interactive term  between  ozone  and  sulfur dioxide.

-------
Both ozone and  sulfur  dioxide were significant  at  the 10 percent level by



themselves.  The interactive  effect was  negative though insignificant.  In



other  regressions  run  but not  reported  the interactive  term  was  never



significant  when  the  air   pollutants   in  question  were  also  entered



individually.  One problem with these regressions was the multicollinearity



which  arose  when a pollution  variable  was  entered  by  itself  and  in



interaction with another.




     It is possible  that  access  to medical  care may reduce the likelihood



that  an individual  will  contract  a CUD.    This would  be the  case,  for



example, if  better access  implied  earlier detection of developing problems



followed by  "preventive maintenance."   To  explore this  possibility,  the



basic  model  was  augmented   with  a  variable   measuring  the  per  capita



concentration  of doctors  in  the  SMSAs  from which the  observations  were



drawn  (PCDOCN).   Equation  (19) presents  the  results for a sample of 7,^04



adults, 7.4  percent  of  whom  (or  567)  reported having CUD.  The coefficient



on  doctors  per capita was negative  but  insignificant  suggesting that the



availability of  medical care, as  measured  here, has  little effect on the



prevalence of CRD.  In future research we intend  to  improve our measures of



access  to medical  care, as well as employ  more  detailed specifications of



the  relationship between  the supply and  demand for  physicians  and  the




consequent effects on chronic morbidity.1




     We also explored the sensitivity of our results  to the inclusion of



data  on ragweed (pollen)  concentrations.   To do so,  the basic model from



equation (9) was run with annual  average  pollen data added.  This reduced



the size of  the sample considerably because pollen data were only available

-------
                                   4-75
for a subset  of  SMSAs.  The results are  presented  in (20).   Ozone was not




significantly associated with CBD in (20) even though it was significant at




the  10  percent  level  in  (9).    The  Introduction  of pollen  had dramatic




effects  on  many  of  the  other  signs   as   well.    Variables   that  were




significant in equation  (9)—age and  education—were no longer significant




In (20).  Variables that were insignificant in (9)—suspended participates,




the   marital    status    dummy,    FAT    and    its   square,   and   annual




temperature—became significant In (20) when pollen was Included.




     A   possible  explanation   for   this   change   may  be   the  special




characteristics of the SMSAs for which pollen data were available—they are




the very largest metropolitan areas in the country.  As such they may have



important characteristics  that differentiate them from smaller SMSAs which



are  not  controlled  with the  other   regressors that  are  included.    To




determine whether  effects attributed  to ozone were • due  instead to pollen,




we replicated (9) on the same set of Individuals used In equation (20).  In




other words, CRD was explained  both with and without pollen data among the




same individuals.  The results  are presented in (21).  Even In the absence




of pollen data, ozone was an insignificant determinant of the likelihood of




CRD in  this  subsample.  Thus, effects attributed to ozone apparently were




not due instead to pollen concentrations.



     This exercise does point out an important aspect of the analysis.  The




estimated  coefficient  on  ozone,  while  robust  in  certain  sensitivity




analyses,  can  change  substantially  across  other  subsets  of  the  data.




Analysis to illustrate the presence of influential observations shows both




that such observations do exist and, when they exist, they tend to be those

-------
                                   U-76
observations where  CRD is  reported.    Of  course, this  is what  one would



expect.    This  warns  one,   however,   that  when  the  sample  is  subset,



previously influential observations can be dropped,  thus  tending to change



the estimated  parameters.   This sensitivity must be  taken into account in



drawing conclusions from these results.



     In reviewing the  analyses  to this point, one very  striking result is




that the smoking variable (number of cigarettes per day)  was not positively



and significantly associated  with CRD  in any of  the  nineteen models.  All



of  these  regressions  were rerun  using the dummy variable CSMOKT1NO) that



distinguished  between  those  who  classified  themselves  as  regular  or



occasional smokers and those who were former smokers  or had never smoked at



all.   This substitution  made no  difference:   current smoking status was



still unrelated to the presence of CRD.



     One  possible  explanation  for this puzzling result  is  that  many of



those  suffering  from CRD  may have stopped  smoking even  though  they once



smoked.   This suggests  the  utility  of subdividing  non-smokers  into two



groups:  those who used to smoke and those who never smoked.  This was done



in  equation (22)  of  Table  U-15  where two dummy  variables  were created



identifying  current  smokers  and former  smokers (never-smokers  are  the



excluded  category).    Both variables  had positive  signs  indicating that



present and former  smokers are more likely than  never-smokers to report a



CRD.   Moreover,  the  coefficient on the  former-smoker dummy  variable was



highly significant  (at nearly the 1  percent  level).   Thus, it would appear



that both  NCIGSDYN  or SMOEI1NO may be  too crude  to  capture the effects of



smoking on CRD.  The results indicate that smoking is strongly related to

-------
4-77


Q.
OC t-
3E — *
o o
•_•* '"S
GC -J
0. w
CL
1/1 if*
r -«
u o

u> C
fV _J
0. w

^ ^
i< f-
C *-•
O -J
L'l C"1
•X nj
a. w

a.
t/> ^
ft- f-
X >-4
0 O
wi a


•
CL
T>
•n
C


o-o p™ <>4 mirv ^rn
t » •• O* O*
^j un in »^ o o o o
1 W •-» W * W * W



•^^ O^ CJ^^» fv|/^
sOCO ^0 ** ^ ^O^
tf •• o* a •
r^O f»O • fi O O
1


f * tt a • o*
rvi -4 m ft o O • <^»
) \«* (N* W » W W
I |

0^ ^»x-» s|.^ fn^
« * • * o • o •
%O nj O **4 O O • ^*
1 ^ 'J' Vrf • **S ^
(

M3^% f- * ^ « O/^
m ^^ »P O «*) O •*) 33
f>o ^f» h»f*> ->4Nr
1 V » • (^ f -3 «
ftO f»'3 0^» OO
t <^/ I W « v> . s/
I

.. ^ ^ ^ ^ ^ 0 ^
(*1O3 «*•-? ^OJ3 <-»Jf»
^»fi ^^O p*^ «^ 0
• • t» a» o»
r-j «. -n — O 0 • <^J
t s^ — 1 *™J * \^ >^
1 1


%a/^ o/^» p**^ o*1^
*f f>-J f i fNJ O — * SO ""I
xf in j^ >"^ m i/% f W * S^ t Srf
'


*O "^ ~O CO •*<• *O f*^ 9)
-- Ul ^0 vt T- ^ 0 t>
^ n-i r-i ^ o 0 O —
I *W ft v^ O S^ • W
*

t—
-" t •* rt ^ O *^ ^ -O I 1 II f^ ^ O ^ OS ^
O*O**»«t • * O * * •
• O O O fi|rg ) f« 0 ^ f« O
1 O
•
1
fi^ ^^ \^ WOWW
• 0
t
1
o « o • •• •* o • o • •«
• OoO|O|fN «OOrj O
1 ^^ * 'w *w SJ" ! ^* O ^V \^
1 O
1
II it -O^N ^r/^ tl II f*-^* co'^ co--*
ii ii *f m as MA t) 11 «^o i/>o o-'o
•srr»i^j-o <*iij*ro^»-j' o*-s if ii o/^ -*/-. T-^»
»l tl \o •« •& T> |t ti co'-J r»-O m^
m 'T* r>| ft O-j -1 <*( ^ i-» ,T>
* t » t vtO***

\^ \^ s^ O ^^ t *^
o
1

II II O *•% O^/^ t) II P->"» u-v^« >r^^
II II 'MO ~t c* (I II u*>«« -T ^* "O ^*
f) O *^ -^f f** u> *^ **~ O 03
O * * t Q • O • • *
• oi<-* .oo^»-^o
v^ W 1 v^ 3 w v*/
O
t

II tl -* i*- f-1- ft o *A OD f— a ^r u^ r~ njaj
^t .Q f4 r»» fi o ^j m — < o fM -o *r -o
— (ft -* fM ( iM O •* IO
W N^ %v« ^> \^O^> W
O
•

> > O t- 2
*t «S fl CJ 3 « O "5
oe nc f*'j_ujuj2u

, "^
^»
w


2^;
o
v^


u> y^
1 0


~~
 f*
<•* n
, »
CT
s^



TO ^N
N. ^
•^ r*
• •
1 o
««/



-1 0
— "^
c>
s^





C»
'^1
do
a.

-------
           4-78
















*
91
ra
e>
in
tft
0

X
i.
a
a
L
>*4
a.
i/>
*
u

u
I**
c
o
L.
j:
v»i

T3
C
It

C
o
4vt
y
r*
**
O
c.
u

CT-
c
•**
.*
o

(^
T5
a>
3
C
.*4
^
C
0
u
w
vn
*^
1

*
*-4
.O
TJ
h—



/*\
<^
rxj
^y



/•%
CD
cv
w







^
f*
*M
W






^
%/l
(VI
SB/







^
IT
"NJ






/^
sf
f*4
W



^
m
r»i
^



^s
f*J
rs*
^







•^
9
Tl
O
r
CL
I/) *•*
a ^
z «
o o
«/) O
!>• -J
Q> W
^
i/> *^
oe »—
3T ^-4
U O
in c
K -J
a ^



0.
wi />
w >-
X *^
<_1 (J»
00 O
(r -J
a, w


^
v> /^
a: »—
X i-
U 0
t/> O
QC «4
a. — i



0.
U> i1^
ae ^
r w
i' •_"
1/1 a
OB N>


a.
IS* /^
oe ^
X "
o o
l^ o
QE ^
a w
a
VO /^
Ci' »—
r «
u o
ui a
nr -J
a.
f» /^
at >-
r —
u o
I/? O
ae _l
CL w

I L.
> m
>
a •
« a
o o
•o
c
•-*

4J O sr A
*s* *n ai IA
O • O I
» O o l*^
w t w


II II
II II









II II
II II








II II
II II










II II
II II







II II
II II





II II
II f 1





t 1 in <•*
II •* CD
f*= fr>
O I
O  ^N     II
          — in
          o   •
           • o
                    O  .
                    o o
 •* ••%    o ^^
 o >o    tn •*

 o   .    o  •
  • »*    o o
 I ^rf     • w
                                                 ^ -»
                                                 (M O
            I ^     f\l ^^    (/S *^
                                       9« *•«

                                       f- O
m i/%     ^ **
o   •     o  v
o o     o o
  • o
  •. » «
          u\ o
Of     Of
o ^*      • ^
O w        w
o
                     i t*l      •  **    O O     1*1 -<
                       v^     I  w     •  «

          o •-
o  •
o e>
                              I
                     • -*    o o
                    I  ^rf      • ^^
                             ^  IN     -• O      .


                                        t f-»     U> I
o o
O w
*r   r~     «-i *^     •• T?

O   •     O  •     o  •
  D f\J      e «rf      • C*

   W      I  W     I  W
  • /^     t/\ *^     O '
  f «j     m o     »< I
  I ^4     (7* O     f^> <
  I   •     C!  •     O
  > O     O O      t •
  t w     O w        •
          o
                    *O  '"N    »•» '
                   O   •
                   O  O
                    2
Z        2        «*
«        «        a
Q^        O        ^^
Z        iiJ        Z

-------
                                   4-79
the presence  of  respiratory disease.   In future research we  intend to go




beyond these limited measures to make use of the tar and nicotine data that



are part of the smoking supplement in the 1979 HIS.




     Differentiating  between never,  former,  and  current  smokers  raises



another interesting possibility:  might  the  effects of ozone  and other air



pollutants on  CRD  differ among individuals  in  these different categories?



To determine  this,  we subdivided a  sample of 7,244 adults into the three



categories:   3,194  never smokers, 1,492 former smokers,  and  2,558 current



smokers.  We then ran our basic regression on each of the three subsamples.



The results,  presented  in equations  (23)-(25), are  quite interesting.   In



equation  (23), based on a  sample  of individuals  who have never smoked,



ozone had  a  very large  positive and highly  significant coefficient.  This



is not  the case in either  (24)  or  (25), which used the  samples of former



smokers and current smokers  respectively.   In neither  of those samples was



the multi-year average  ozone  concentration significantly associated with



CRD.   Sulfur  dioxide was significantly  associated with CRD  among former



smokers as  it was among  never  smokers,  but it was  not significant in the



current smokers regression.



     The  results  of  (23)-(25)  suggest  an  intriguing  possibility:    if



individuals elect to begin and continue smoking, they may so increase their



likelihood  of developing CRD  that  the marginal  effect of  exposure  to



ambient  ozone  pollution  may  be  very  small  or  non-existent.    Among



individuals   who  do   not  smoke,  however,   ozone   is   positively  and




significantly associated with the likelihood of CRD in the models estimated



above.

-------
                                   4-80
     To  explore  the  sensitivity  of  these  findings,  we  replicated  the

experiment in (23)-(25) in a model that  includes  five pollutants.  In this

model,  first  presented  in  equation   (10),   ozone  was  positively  but

insignificantly related to adult  CRD.   However, when  separate models were

estimated for never,  former  and current smokers, ozone  was  positively and

significantly associated with CRD among the never-smokers (Equation (26)),

as it was  in the three-pollutant specification.  Among  former and present

smokers,  however, the  association  was not  a  significant   one.   Sulfur

dioxide was  positively and  significantly  associated with CRD among never

and former smokers, while total  suspended  particulates were  positively and

significantly related  to  CRD among  current smokers.  In all, (26)  - (28)

strengthen the case that long-term exposure to  ozone may be  most likely to
                        »
result in CRD among non-smokers.

     Throughout the analysis of  CRD, the same  basic set of  regresaors was

used.  Air pollution was summarized in different ways and sev«sral Different

measures  of  smoking  status  were explored, but little  was  done  with the

other independent variables  even though a number  of them were  seldom if

ever significantly associated  with  the  likelihood  of CRD.   Dropping them

because they did not perform as expected could have biased the coefficients

of the ozone and other air pollution variables of interest.

     As a final experiment in the analysis  of adult CRD, however, a "lean"

version of the  basic  model  was tested.  This version excluded a number of

the  variables  that were  never  significantly associated with CRD  in the

earlier analysis.  The results from this version of the model are presented

in  equation  (29) in  Table  4-15.   They do not differ  substantially from

-------
                                   U-81
earlier results.  The multi-year  annual average hourly ozone concentration



was positively associated with the  likelihood of CRD, at between the 5 and



10 percent level.   Neither suspended particulates  nor  sulfur dioxide were



significant.  Income was negatively and quite significantly associated with



CRD, while age was positive and significant in the same equation.








      U.3.2  Children



     The  analysis  of ozone and chronic  respiratory disease among children



is  much more brief  than that for  adults.   This is because the analysis



revealed no relationship between ozone and CRD in children and the approach



in  this  study has  been to confine  sensitivity analyses primarily to those



areas  where a  positive  and  statistically  significant  ozone  effect  was



identified.    Even  had  such effects   been   identified,  the  sensitivity



analysis would have been more limited than that above because less data are



available  for  children.    For   instance, information about  almost  all



children  was  provided  by  a parent   or   other  related  adult—thus,  no



"self-respondents" analysis could be conducted.  Also, no smoking data were



available  for  those  below  the   age  of  17  nor  could  accurate  household



smoking  profiles be  constructed.   This limits  the  variations on the basic



model that could be considered.



     The results of  the  analysis  of CRD in children are presented in Table



U-16.   Equation  (30) is  based  on  a sample of 8,483 children  aged 16 and



below.   Of this  group  516 had CRD.   The corresponding figures for equation



(31) are  6,886 and 429 respectively.   Ozone  and other air pollutants were



measured by  1979 average  concentrations  in the first two models.  Although

-------
4-82























c
3>
k
^
•ri
•ft
£
U
e
o
E
«
»
V)
u*

O
X
o
«•>
m
k
.•4
a

•
ae

u
^4
C
o
k
•
•o
*<
i
•f

•
•*
.a
•
*-





in
*o
w



/^
*
m
w



^
m
w




w








%

*^
«
•o
o
z
0.
V) <•%
oe >-
Z M
k) O
i/» a
(r _)
a. w
a.
i/» <->
CC w-
z fc*
u o
ut a
ae _j
& w
a.
cc b.
Z M
U O
1/1 a
ff t
G. W
a.
ee h.
Z M
U* (J
i/i n
ee _j
a. w






a
OC K.
Z ta
v> a
»
O (>9
vi a
cr _j
a, w



k •
t k
> IB
>
•
a •
• a
o •
•o
e
!-•
<•» ** ii ii ii ii ii
r- f- II II II ii ii
&• **
t i
m «v
I w



m « II ii (I ii II
** •* ii ii il ii ii
« •
IV IV
1 w

• A 11 II El II 11
^ It II II II II
^ ff* t 1 II II |l II
<* •*
* t
1 yy


•*> « ii ii ii ii ii
•0 aj Ii || || || ||
•* *4
• •
n •f
1 w







mi>. iva oe> <- m oiv « ooci iv«o
m m«^ a)iv  u z
m^ **^» cr*^ ii in'** *»/N o s\
iv iv no mo i i o>iv o ^>o «^o u>i^ !*>•*
• • o • o • e • o • ii
09MOOOO OfH •» IV
f|^^^O^'O^•' l««r s/ •**
III |
1 1

0»*> O'* .-« || II **% «/^
^>O ivi*^ in in || ii *»<* ^^
• •OlOl III!
in«4oooo « N
•4 W • ^ I X^ ^ ^
1

rs) »M f- ff* *40O if» iTt f«»«O f) *^ ^
coo* »**n ^es ^s ^ tf\*n m**
• t O • O « • • C • t*
(7^ O O O O *•« fNJO** P4
1 ' 1

*Of*< ^.^ O**-^ || I) || Ite^o
^ O 05 00 09 sr |1 It II O -T
•M O \T <^J ^) «^ ft f*f
O « O * O • e*
• O O *^ O ^ csj
W t W t W W
1






II II II II II II ^ ^
II II II 1 I II II O ^
-r -o
" i^
w





II II II II II II -o /-.
II II II II II II IT- en
m CD
i i
IV
w







> » > > > o o
« < « « « a. a>
ff ui ee oe K « •« .
« 0 « « « 0, 3
Z Z Z Z Z «rt Ul
< 
-------
4-83





















c
II
L.
•o
X
X
c.
u
o
c
o
6


l/i
ffl
0>
VI
0

o
•*
A
L.
a
«i
4
U
C
0

r*
•o
•
c
X
c
0
w
„
t
^

o>
•<
£1
a
*•


*A
fO
w




^
^
f^
w




<*>
m
w








/>s
r^
m
w








^>
x
^







o
m







w

•^
•
^
o
X
A

fie ^
X. >-<
u o
v\ a
oe .j
a. w

a.
or /*
X H-
u w
M 13
e. -i

0.
n /^
ac t-
X M
U O
->
IJ U
v> a
Ov w





o.
X **4
O C9
to a
cc _l




b t
> Q
>»
•
a. •
« a
o »
^
c
<->

IS M in NT
ft O •

in M o 09
« CD •* *•
• . Of
(n o O
O
o
t



CD cr- m IM
m r* «-i IM
• 1 0 •
m oo
o
o
•




•o 
»* X
x: a
X O
UJ Z
I/I ••*

mo r- in  0 I
i m >x o o
w w • w
1



o in o c" —on
^r ^ IM ^ f* IM
O 1 O' • O 1
• n • *• o <-i
1


m x IM o> m IM
i • o • o •
o m t ** o ••*
I







f- *> vO l-< IM /•»
IM m in •* IM o
** -*• ^4 p^ in ^
o t ot o •
t m I •* O ••
1





^-  O
m *M c* IM •* o
O • 0 • O •
• m ox o o
•
I






3? 3?
< <
a. o
>n y m
13 Ul CC
< r- 0,

CD M  x -f ea







Z IM
« X
o
X L.
X 0

X U. Z

-------
                                   4-8U
positive  in  both equations,  ozone was  not significantly  related  to  the



likelihood of CRD in  either specification of the model.  Neither were the



other air pollutants,  with the exception of total suspended  particulates in



equation  (3D where it had an unexpected sign.   Nitrogen dioxide, which was



sometimes positively and significantly related  to acute respiratory illness



among children in the analysis above,  was  positively but not significantly



associated with CRD in equation (31).



     The  significant determinants of CRD in children were race (whites are



more likely  than blacks to have CRD)  sex,  and  age.  Since  we are dealing



with children in  (30) and  (3D, the significance of the coefficient on the



age variable is to be  expected—the very youngest children would in general



not have had sufficient time to develop or manifest CRD.



     In  equations  (32)  and  (33),  we  replicate these  results  using  air




pollution averaged  over as  many  of the six previous years as  there were



valid annual  average  air pollution data.   For  the  very youngest children,



the data  In  these  two  regressions  may characterize  concentrations  at  the



monitors  nearest their homes over their entire lifetime (if  they have lived



in the same location reported  in  the  HIS interview).  Using the multi>-year



averaged  pollution  data  made little difference  in the findings about ozone



and CRD in children.  In equation (32),  however, sulfur dioxide entered as




positively and  significantly related  to CRD.   Race,  sex, and age remained



important explanatory variables under the alternative specifications.



     Finally,  we  ran   two  regressions   which  included   PGASPIPD,  the



probability that the child came from a household where natural gas was used



as the  cooking fuel.   These  are  reported  in equations  (31*) and  (35) •   In

-------
                                   4-85
neither  were  any  of  the   air   pollutants  positively  and  significantly



associated  with CRD  at the  5  percent  level.    In (35),  however,  carbon




monoxide was  significant  at the  10  percent level, a finding  for which we



have no explanation.



     In Equation  (34), which includes  three pollutants averaged over the



period 1974-1979, the gas stove variable was positively associated with CRD



in  children at the  10 percent  level.   When carbon monoxide  and nitrogen



dioxide were  included in  (35),  however, it  was  no longer  significant.  In



(34) and (35)  race,  sex and age  remained significantly associated with CRD




in the children we examined.



     To  reiterate,  we found  no  association  between  ozone  and  chronic



respiratory  disease in children.   Admittedly,  our search was  a somewhat



limited one,  but  only because our basic model—which  often suggested some



associations  in our examinations of acute  and  chronic respiratory disease



in adults—gave us no reason to believe such a search would be fruitful.








4.4  Cardiovascular and Other Chronic Diseases



     To this point our analysis has been confined to acute morbidity due to



respiratory and  other  causes,   and  also  to chronic  respiratory disease.



However, air pollution may increase the  likelihood  of other  diseases as



well, including disease of  the cardiovascular system.    Our final analysis



of  chronic  illness concerns cardiovascular and other diseases.



     Using  a sample of 7,244 adults,  of whom  733 (10.6 percent) reported



having  chronic cardiovascular  disease  (CCD),  we estimated the  models in



equations  (36) -  (38) in Table 4-17.  In none of  the three models were any

-------
                                   i-86
Table 4-17.  Analysis of Chronic Cardiovascular Disease
Model #
Dependent Var.
Independent Var.
INTERCEPT
03ANARAV
SPANGEAV
S2ANARAV
COANRAV
N2ANARAV
RACEW1BO
SEXM1FO
MARY 1 NO
INCOMCON
SMOKY 1 NO
SCURRENT
SFORMER
FAT
(36)
PRSCHRSP
(LOGIT)
-6.884
(6.12)
13.174
(1.61)
.0000370
(0.00)
.000741
(0.24)
--
—
-.515
(4.26)
-.200
(2.19)
.383
-(3.98)
.0000293
(4.51)
.132
(1.39)

— «=
1.404
(2.37)
(37)
PRSCHRSP
(LOGIT)
-7.713
(5.67)
9.431
(0.98)
-.000575
(0.14)
.00148
(0.40)
.0532
(0.86)
-.000304
(0.10)
-.471
(3.49)
-.286
(2.84)
.326
(3.10)
-.0000216
(3.72)
.209
(2.01)
—
—
1.444
(2.22)
(38)
PRSCHRSP
(LOGIT)
-6.948
(6.17)
13.308
(1.62)
-.0000693
(0.00)
.000635
(0.20)
* .__
..—
-.531
(4.38)
-.291
(3.07)
.353
(3.65)
-.0000239
(4.49)
.__
.291
(2.76)
.396
(3.68)
1.418
(2.39)

-------
Table 4-17 continued
                                   4-87
Model #
Dependent Var.
Independent Var.
FATSQ
AGE
AGESQ
EDCOMCON
TEMPAN
PRECAN
HUMID AN
F or X2
N
(36)
PRSCHRSP
(LOG IT)
-.0852
(0.78)
—
.000549
(21.45)
-.0308
(2.32)
.00485
(0.59)
.00585
(1.75)
.00862
(1.01)
949.12
7,244
(37)
PRSCHRSP
(LOGIT)
-.0810
(0.67)
—
.000564
(19.93)
-.0214
(1.45)
.0122
(1.05)
.00587
(1.39)
.0101
(1.02)
773.78
6,008
(38)
PRSCHRSP
(LOGIT)
-.0869
(0.79)
—
.000547
(21.21)
-.0327
(2.46)
.00480
(0.58)
.00590
(1.77)
.00852
(0.99)
962.52
7,244

-------
                                   4-88
of the air pollutants positively and  significantly associated with CCD.  A



number of the  other independent  variables had the  expected signs and were




significant.    For  example,  the  dummy  variable  for  races  was  highly



significant and  is  consistent with earlier  findings that  blacks are more



likely  than  whites  to  suffer  from CCD.    Income  was  negatively  and



significantly associated with CCD,  as  was education.  Both findings confirm



£  priori expectations,  as  does  the  finding  that  the  likelihood  of  CCD



varied  directly with  weight.   The  square  of  age was  by  far  the  most




significant explanatory variable.  According to (36) - (38) women are more




likely  than men to  report  CCD,   as are  married  as  opposed  to  single




individuals.



     The  crude  dummy  variable  indicating  current  smoking  status  was



significant at  the  5 percent level in  (37), although it  was  not in (36).



As expected, it  indicates that  smoking  is  positively related  to CCD.   An



additional regression identical to (36)  was  run substituting for SMOKY. 1N0,



the pair  of dummy variables  distinguishing both current  as well as former



smokers.  The  results are  reported  in (38).   Differentiating between those



who never smoked and those who once  smoked  but no longer do makes a great



difference.  This model indicates that both current and former smokers have



a  significantly  greater risk of  CCD than non-smokers  (p  < .01).  There is



little change  in the other signs where SCURRENT and SFORMER were included



as possible determinants of CCD.



     This analysis  of CCD should be  viewed  as preliminary.   Importantly,




the  regressions  presented  here  contain  no data  on  individuals' alcohol



consumption (because  no data were  available from  the  HIS).  This omission

-------
                                   4-89
may  be  an  important  one  in  light  of  previous  findings  that CCD  is



positively and  significantly related to  alcohol.     If alcohol consumption



does not  vary systematically with air pollution,  however, the coefficients



on ozone and the other pollutants will not be biased.



     Finally, based  on the  recommendations  of  a number  of  individuals we



contacted  at  the outset of  our study,1^ we explored  in  a preliminary way



for  possible  associations  between ozone  and  other  air  pollutants  and



several  types  of acute   and  chronic illness.   The  acute  illnesses  we



explored  included those   of  the  eyes  and  ears  as  well  as  a  group  of



infectious diseases.   The chronic  diseases  included neoplasms,  mental and



nervous  disorders,  diseases of  the  digestive  tract  and  several  other



illnesses lumped together  under "other symptoms and  ill-defined conditions"



in  the  HIS   diagnostic recede.    Using  our   "basic  model"  we  found  no



association  between   ozone and  any of  the  acute or  chronic illnesses we



examined.  These results  are not  presented here but are  available from the



authors.







Analysis  of Multicollinearity



     Multicollinearity  in  the  linear  model  has  been defined as  the



situation where "the rank of the  regressor matrix  X is less  than K, the



number  of regressors," and  "near  multicollinearity" to  be  the  case where



typically "one  of the  regressors  is  'almost'  a  linear combination of the


                        T                           1 ft
others....in this  case XX is 'almost' singular..."     The implications of



multicollinearity  and near  multicollinearity  for parameter  estimation in



the linear,  nonlinear, and multinomial  logit models are  fairly well known

-------
                                   U-90
and will  not be  reviewed  here.   Rather,  in  what  follows we  discuss the



analyses  of  collinearity  we have  undertaken  in the  present  context  of



assessing the effects of air pollution on human health.



     First, we examine  the  question  of  correlation between the independent



variables used in the models estimated above.  Then, we describe analytical



tools used to detect what  is known as harmful or degrading collinearity.



We  concentrate  on  collinearity  in  the  linear  model since  most  of the



estimation done in  this project uses ordinary least  squares  in the linear



framework.   Following  this, we examine  the empirical  diagnostic results



emanating  from   a   subset   of  the  models   estimated,   and  assess  the




implications of  these  results  for the interpretation of  the coefficients



estimated in this report.



     The reasons  underlying this  exercise are several.   Because the world



is  not  orthogonal—>that  is,  not  all  variables can  be  presumed  to  be



independent  of  one  another—-the  attempt  to estimate  partial  derivatives



will often  be  fraught  with  error.  The diagnostics  referred to above are



invoked to indicate the extent to which our estimates are confounded by the



presence  of  significant covariation  between  and among linear combinations



of  variables  in the design  matrix,  and to  identify  what these covariates



are.   The  diagnostics  we  use are  particularly helpful in  this  respect.



Whereas  simple  correlation  coefficients  allow  one  to  see  where  the



covariance   between  two   variables  might   prove  harmful  to  parameter



estimation, this is not the  case when the covariance arises due to a linear



combination  of  more  than  two  variables.  For  example,  consider  a linear



model of the form

-------
                                   4-91
                      y = a
or
                      y = XB + e

Assvme we are given the data

                   »                «
                   1   2   2.1  0.1
                   1   7   7.1  0.1
                   1  14  14.0  0.1
              X=   1   3   2.9  0.1
                   1   6   0.1  5.9
                   1   5   0.1  5.1
                   1   2   0.1  2.1
                   1   6   0.1  6.1
                   f                 •
The simple correlation matrix is
                  R=
1.00  .83

     1.00
                                                  and  s..  having the usual
                                    -.08

                                    -.62

                                    1.00

where r^  =  s^/Cs^), i,j £{1,  2, 3},  the

definitions as sample standard deviations and covariances.

     Upon  examination,  one might be  convinced  that X   and  X- are closely

correlated (r*12=  .83),  but that the  correlation between X. and  X- is not

very strong (r  = -.08).  However,  consider X4= X£ + X_.  It turns out that

P14=  .999, I.e.  X1 is  almost  a perfect linear  combination of X_ and X_.

Due to this nearly perfect collinearity, parameter estimates will likely be

seriously degraded in estimation of models such as that described above.

     There   is   another   rationale   for   exploring   these   issues   of

multicollinearity.    There  has been  considerable  concern  expressed  in

previous   epidemiological   studies   about   the   problems  in  interpreting

coefficient estimates on individual pollution variables when such variables

-------
                                   U-92
are  correlated  with each  other, with  meteorological variables,  and with



other explanatory variables.   Even in work where much effort is devoted to



examining  multicollinearity among  pollution variables,  concern  is  often



chiefly with  simple correlations between  two  variables.   Inasmuch  as the




primary objective of such studies has  been to  assess the  influence of each



explanatory variable individually on health status, such concern is rightly



placed.




     The  bulk of   the  interest in collinearity  problems  in  statistical



studies like the present one has focused on the covariance of the pollution



measures employed and on the related point of covariances  between and among



pollution  summary  measures  and  meteorological  variables.   In  areas  where



some ambient  pollution  levels  are  high, others  tend to be high also.  For



instance,  because  ozone and  other  photochemical  oxidants  result  in part



from vehicular  emissions  of hydrocarbons  and nitrogen oxides,  oxone tends



to  be  correlated  with  carbon monoxide,  another mobile  source pollutant.



Similarly,  the  nature  of   the  physical world  is  such  that when certain



meteorological  conditions  prevail,  so  too  do  levels of  some  pollutants.



For  instance, sunlight  and  heat tend to "cook"  hydrocarbons and oxides of



nitrogen  to  form   atmospheric  ozone.    Thus,  ozone  and temperature  are



positively correlated.




     We  have  earlier  examined the  bivariate   correlations   between  the



explanatory variables used  in our  analysis.   Recall that  it was shown that



certain   of   varibles   tend  to  be   almost  perfectly   correlated  (e.g.




rAGE AGESQ='^2)»  other pairs  quite  significantly  but  not so  perfectly



correlated  (e.g. **      AVMAXTMP=e5^'  wnile  other variables  tended to be

-------
                                   4-93
virtually orthogonal  (e.g.  r*03NR(^1 AGE=-.003).  We  now turn  our  focus to

more sophisticated collinearity diagnostics, reemphasizing that examination

of the simple correlation coefficients can be quite misleading in assessing

the   potentially  harmful   effects   of  multicollinearity   on  parameter

estimation.20

     In the presence of severe collinearity, the least squares estimator of
   /V
3, 8 = (XTXr1XTy and the estimated covariance matrix, s2(XTX)"1 tend to be

numerically unstable.   In  both cases,  this  is due  to the fact  that XTX

tends  towards  singularity as  the  severity of  the  collinearity increases,

with   the   determinant   (XTX)  tending  to   vanish.     Such  collinearity

corresponds to  the  existence of at  least one  small  eigenvalue  of  the XTX

matrix,  with  the  numerical-analytic  concept  .of   condition  numbers  or

condition  indexes employed  to  indicate  just  what   constitutes a  "small"

eigenvalue.   Simply  put, large condition  indexes   are representative of

small  eigenvalues.   It  is important to  note  that  in the linear model, the

problem  of  collinearity  is  manifest  solely in the  X matrix  without  any

reference   to   the   error   structure   or   dependent   variables   under

consideration.

     There  is no statistical basis on which  one can judge when harmful or

degrading collinearity is present even given the magnitude of the condition

indexes.    Experimental  experience  has  shown, however,  that  condition

indexes  on  the  order of  5  to  10  tend  to be  associated  with weak linear

dependencies among the  columns of  X, whereas  indexes of 30 and above tend

to  be associated  with moderate to  strong  (i.e.  harmful and  degrading)

collinearity.

-------
                                   4-94
     The effects of collinearity have also been examined in a more specific



sense, that  is, via the  decomposition of  the variance of  each estimated



parameter.   In this approach,  the estimated  variance of each regression



coefficient is  decomposed  into a  sum  of  terms, each  term associated with



one singular  value (the singular  values  are the positive square roots of

                                        /\      /\


the eigenvalues).   For each component  3 .of S  ,   it has been shown that

                  P                       K

var( ek)  _   a2 ^ (v   2/  Uj2).  Here, [v  ] are the typical elements of V


                                                                   T
where V results  from the  singular value decomposition of  X, X=DD7 , where

                            /s.

it  can  be shown  that  var( 3)  = cr2(XTX)"1  =   cr2VD"2VT.  U,  is  the J-th

                                                             J  P

singular value.   If one defines   4>   = (vk- / V  ) and  *  k  = j=i*kj» tnen



the k,j-th variance decomposition  proportion —  i.e.  the percentage of var



(BO  associated  with   the  j-th  component  of  its   decomposition  —  is



calculated as   H   =  (4>jc1/*k).    (These  collinearity diagnostics  can be



calculated In the case  of  the linear model in SAS's PROC REG  specifying the



COLLIN option in the MODEL statement.)



     Strictly  speaking,  the  analytics  described  above  are  not  fully



appropriate  in the case  of  the  binomial  logit  model.21    In  the  linear



model, the problem of collinearity was shown above to be fully attributable



to  characteristics  of the  X  matrix,  specifically,  significant  linear



relationships  among the columns  of X.  Here,  the  covariance matrix of the


                                                          T  —' 1
estimated parameters is merely a scalar multiple of the (X X)   matrix.  In



the  binomial  —  or,  more  generally, multinomial  — logit  model,  the



estimated  covariance matrix equals  J   ,  where J,  the information matrix,



depends not  only on the data  matrix X but  also on the estimated logistic



distribution   function F (X (3).     It   has  been   demonstrated  that  the

-------
                                   4-95
fundamentals  of the  regression diagnostics  for  the linear  case  can be



applied  to  the analysis  of collinearity in  the logit  case,  although one



must work with the information matrix instead of the  (X-'-X)'1 matrix.



     Although  we  have  almost  completed  work on  the software  capable of



performing the collinearity diagnostics in  the  logit case, we have not yet



been  able to  perform  the  necessary  benchmark tests.    Because  logistic



estimation has  been important  in  many facets  of this  study,  however, we



have  performed collinearity  diagnostics  on  the  X  matrixes  used  in the



binomial logit models  as  if_ such models  were estimated by OLS with binary



dependent variables.   Although  different  results would be forthcoming were




the  estimated  logit   covariance  matrixes  to be  decomposed,  the present



exercise is nonetheless illustrative.   Of course,  for the models  estimated



by  OLS,   the  collinearity  diagnostics   described  above  are  entirely



appropriate.



     We  now  turn  to  a  discusssion  of  the  collinearity  diagnostics as



applied  to twelve  of  the  models  estimated  in our  project  and  discussed



earlier in this chapter.   This  is  followed by a tabulation of the full set




of  collinearity   diagnostics   produced  by  SAS:   eigenvalues,   condition



indexes, and the variance proportions matrix.



     The most  important result in  the context  of  the present analysis is



that   the  ozone   variable—however  summarized   in   the   models    under



consideration—cannot  be  said  to be a partner  with  any other variable or



set  of variables  in  harmful  collinearity.   The  criterion on  which  this



conclusion is  based  is the  following:   given an associated condition  index



of  30  Or greater,  the proportions of  the variances of the ozone  parameter

-------
                                   4-96
and that  of at  least  one other  variable not be  greater than  0.50.   The



single exception to  this  occurs  in Acute Model  26 where the ozone measure



and the ozone-temperature interactive term measure appear to be involved in



some  potentially confounding collinearity.   This  would  be  expected given



the nature of the variates.



     Equally  noteworthy is  that,  by the  same  criterion,  all  pollution



measures  in each of the  twelve models  (except  Acute Model  26) withstand



scrutiny  for   involvement  in   harmful   collinearity,   as  do   all  the



climatological measures.   Of course, the criterion  established  here is an



ad hoc one  for the reasons discussed earlier.  However,  it is a reasonable



and often-utilized one.



     What  would  by  our standards  be termed harmful collinearity exists



where one would  expect  it, i.e.,  among functions of the same variable.  In



many of the models estimated over the course of  this study, this refers to



the (AGE, AGESQ) and (FAT, FATSQ) pairs.   The inclusion of both linear and




quadratic   terms  of   the   same variable  is  often  found  to  be   a  major



contributor to  collinearity.  This is the  case  in  the present analysis.



However,  a_ priori restrictions  on  the  specification (i.e.  "leaving out"



either  the  linear or  quadratic  term)  constitutes  the  introduction  of  a



prior  restriction  on   the  estimating equation,  which can  itself  have



undesirable properties.   Testing  for both linear  and quadratic  effects in



the cases of AGE and  FAT  seems  quite plausible.   Therefore,, one  is left



with a tradeoff  whose resolution is beyond the scope of  the present effort.



We  have  little to say  about the involvement of other  variates  in harmful



collinearity.    It  is  generally  true   that the intercept  is   involved

-------
in the colllnearity with FAT and FATSQ, suggesting a strong relationship of



the form FAT = c +  b*FATSQ  (plus some error).   Otherwise, there appears to



be little  degrading collinearity present  in the models  scrutinized  here.




In  further  research,  we  hope  to  analyze the  logit  models  in  a  more



sophisticated manner,  and  will  perhaps have different conclusions  to draw




after such an examination.

-------
o
o
    • r   .••••.••••••
    Q3  aoooaoooaooa<
    a. z



    o^  o.*«* o o> a «»«i •• o 4 «•«
    «*u  a o o a «• a » » W a a*. <

    Ota   •  •••••••••••
    oo.  aoaoaaoooa»o«



    rz  o — — o»« — ow o»» «





    a.



    OO  e»o — >M —n» a a o> MI«M <



    ou  aoooaaaoooaa«
    & a
      ui  	           ______

    za  a —«••« —aa««.«4«

    22  •2J?SSy2S2-!llS;

    ac «   ............
    o    a-a*oaa«aaoaa«





















    ZZ  »—>.*•*»• MO> «•»•« —


    *-r  aa«««waMa-»aaa
    «o   ••••••»••••».

    a. z



    oz  —.««»«> a »» «,»-••> .»o «-
    .•..  a^^^ ««VOM ^ .«« a> o a <

    oeot   ..............




    Zo  »»in« »*• »* «i M^«> a « vt a <
    o *b  ^^«^ M^»«« 4>M «r «»«»oo tf« <

    t> x
    oe »
    0 W
    a, •


    z o
    OB


    ac w
    ow



    z »
    o «


    at z
    o •«
    •» **


    z »



    •E Z   ...««c««"»«.c.«.




»  Z*  •».»»•••.•»•.•••«•, a ••»••. we



0  ITz  "*"*"* "***I^°.***
at  o •«  aoo oaaaoaaaaaoa

      O
w

Z  ow  aoa aao ooo a o «-o •> m
«  «u  oo a ooaooa oaaa «><•


|  ||  32^jj22222I-22«^














I  ^
o    w

••    -*

      >  «•» «»rf>IV«
>>    z  ••• *Mm •
                                                             4-98
                     —«Jt>» « W •«•»«» O—><

-------
                    2 Z
                    O <
                   ex ac
                   o «.
                           a o o »>• o«< «»• o >«» t»
                   O      0000««0««
                      5»    OO O «•*»
                      ^    00*000
                   o      oo<
                                ..oovoo*
                           OOOOOOO-WIM
                      SO     ••••••
                      U    OO«OOO
  u
 •^

  o
  (-,

'u
  0)
 T3
  O
t-t    omo«ooo<
deaf     •••••••
O 141    OOOOOOO<
                    O«   O O O —«A1 Ml «••«
               K    O 4   O<
               0    —«   O.
               a.    «- •»   01
               o    oe z    •
               K    o •»   o<
               a,    a. m
                      o
U   Z ft.   00 00 O O 0 — -«0 O •« —0»
z   OKI   oo oo o o o oo o —-• o » «•
«   MU   OOOOOOOOOOOOO••
»   » c   oooo o o ooo oooorar*
K   OCUJ    ••••••••••••••«
«   at-   ooooooooooooooo
                   O IU   O » IMOI
                   •• o   o ni M in i
                   ». z    «...
u   e
«   z
H.   a
•»   w
Q
z
19
                      ui    KI o —
                      9    -O « 1MI»
                                                                                4-99
                                                                                             OC  V     • •
                                                                                             03   00.
                                                                                             &  £
                                                                               z z   « m>
                                                                               o •*   o o.
                                                                               » u   o o <
                                                                               X X
                                                                               a &
                                                                                                       oooooooooootnmo
                                                                               z z   •»»«
                                                                               o a   o<»«•% —
                                                                               «. • z   oooo o<
                                                                               so    • •  •  • •
                                                                               O W   O *» O O O '
                                                                               0.0
                                                                                                                _9 •« 0« -O — O — ,/<
                                                                                                        IW*> •« — «•«•• «-M O -«
                                                                                             o      ooo«oooo<
                                                                                             »,

                                                                                             z »-   oofcoo —««<
                                                                                              IO   OO«««M OO«*«> • O — ^M<
                                                                              ou   •oo>«o«ooo,«o«
                                                             (M


                                                              U
                                                             JS
                                                             u
                                                             0)
                                                             TJ
                                                             O
                                                                              •• r    on

                                                                              o w    o«
                                                                                             •• <*
                                                                                             • z
                                                                                             *"
                                                                                            >OMOOfll»«MrtOOO
                                                                                            kOOOOOO«»<0OOOO
                                                                                            • ••••••••••••
                                                                                            toooooooooooo
                                                                                        ff    CM
                                                                                        <    O »•
                                                                                        •-    o
                                                                                        z
                                                                                        <•

-------
                                                                                4-100
C
O
           out    — e M«t «•>«•«•»•« o •« « « o « n*
           2C    e. o o o •• o
                   e o o o o o o »««v * •• o o»»e
           ••*•    eoeeeoooooo
                             oeaeooeeoee
                     o o o o o o ooc o o c o e M
                               oooeccccoco
                   00000000000000000
           OO    e e a — — «•-<••*«•—•«««•>*»«•!*
                                  ««o* c« -»o e o c
                                            » o o e o o o
           •»—    c. o o — c tt *»o
           ^x    o c. o o <• e
                        o«oooeooeoeeeoc
           »-<«    cooocooooo-«a— o —
           2>    — — e «* — « «^- ut^. ,-Atn oo> ni^>   •- u>    o « -* a o c a c e c c. a
                                                               O _i    CQOOOCOCOOCC.OOOO*
            - o    « c a e c o oo o oo o —•»«•» oo   ao
           az     ..«..«.»«»».»...»      a
      z    a a.                                                22   e oo« — ^c- »«
      O       i/>                                                O«eeeeeeese
      —                                                        ~Ccooeoecac
      ^    Z>    <•«   »• ••   e e o o e a e  n — «•« «    •• i*   e e e o o e o e c « — & u
      Z    &w   00000000«000««f»0«tl*t    Ol tt    «.«••«.  ..ceoc.
      «   •» o   ooaoooo-oooaoeee«nt    o «L   ooeoeoeooeoeo
      •-   »-  or   c e e o « o oo o oo a oc of«r»    a,                   —•-•«»»»» w
      (C       a.  Z                                                a, z   e oa «> o « »— > o •>»> ^ m
           O m«»>.(vib«-


                            o OB ««•<•>••• m At _» Q o o    -» o   o <• ««• e « o o —o o e e e o a o
      >•      z   ««»»^.*»«««»»««»"»oooooo^    *-wi   o««oeoooooooooooo

      —      o      o o o o ooo o oo o ao oo o    o •«   o o e e oo o a oo o e o o o oo
      «      •—                                          o    a, u
      «      Ml                                           •      ^

      Z      «   •« ni « * lAOf^OO O»»
-------
                                                                                 4-101
             za   o o«*ow o«»« «* oo«t o« M«*
             cut   ooooo e o«« «u**o o •»••••• o0>
             — *-   o o o QO o oo o oe o o o o«»o
             C
             a.
             3^-   o o o eo o o •« »
             O«   o o o o o o o o o o c o a o o «—
             •-%.   o o a e e oo e o e o e e o o
             ^      a oooc o a e e ae e e o e
             o
             22    *»•»!••»«•» *»« •* 0«« f-« «•»•• «•«
             OO    o o — .«-. o —*••«•« — «< — o o .-<=
             «•      OOOO^O*I^^-**OOOOOOO
                    ooooa O«*»M oo •» o o oo«»
             aa
                    o
             ?o   M<0M«i« *»—«•* *fti«t * o a
             oz    —«•—*»«*-• o * c *» e o o
             • a
             o «
             a. a
             20   »l>t«l«l « Al~«t « »••• Orfl x
             Oh.   «0k o«> «. r» o « M oo •* oo
             i—•»   a o • OKI o o *• *> •« c o o o o o^
             ^X   a o« oo o o o o oo o oo o oc
             ax    .«««««»»««»••«.»•
                    eeoeeoeeoeeooeeoc
             o«   o o o o«« a»M» P> » a «« « *•
             — —   c e o c a .« M\ •»« F» G«v««^nnQO
             •—«    O O O QO « MM O •«!* O « O« OO
             »• «   00 OOO O O O O O

O O»>« OO U az .«.««».«..«•••••• C « eoooeooeeeeoeooao * i^— «n<— «o>— »-t»*i»^-«<««c ^ — ooooooooooec.c.ani«a o« ocaoao«oe«n<*«niOMt<: a I -««».••».«.....». —a o o o o o f»— o o — o e — o o eo oz e o o ooe oo o o o e e o e o o >>« ocoacooeoco»«cooo a I ac2 «.«...«....»..«.. wi c « o o o oo o o o a e o o oo o oo z a, « 22 «e ««M«I O c o«eo~ — c-* — ~ u o e o oo»» >- ZC «••••* *»•« •« M « o «»» ••» »• «•• ^ u» o o o ooo o o o e e — o « o ~ « oo. oo»«oui«eo>o-»o« —•• aca •.»«•»••»».««««». c "•— o o o c o o (M o c c ««<•••« ^- o ca e o o ooo o e e oo o e o e a o a. ^o. ooaoeoeooo«ooo««ioo a. o aui ................. • 1 o o e o o o o o o o o o o o o — o «* -= o o oo oo o o o o oo o o o oo * Of e o o ooo o o o o o e o o o o o •- H- a, * a ut ................. •« o^- oeooeooooeceoceoo > 0.2 22 «o _>_««» O««KI «««« -.o *-Z oeoooooaaoe««-ec.oo arc ................. 2x 0(«»-o<»»«<«*»n^. C em««c»«m<«itfi9O>MO«>Mtai9 w M •-•* -• •* «*» K> «V >» • » C — •« « IM W » — -* 22 C K 4 » ••*• O ^ Mt» m tfl • o> — •> e in n oA<>«oooo Z «»>i«««i«««.«.«oecoooo* KO eeo»«ooooeooeooeo •- kj «••••»••••. ..«eo 0[ « .»••»«.«•«.«....« — o oaooooooeeoooooo c eceoooeoooeccoeeo a » a a, _t C -» X o =i w z


-------
                                                                              4-102
           f, ».   OOOO O OO •* M—OO O—4
           o<   ooo o oo oo o oo o o e o a at »
           •« ^   oooooooooooooooauiA



           O      OOOOOOOOOOOOOOOOOO

           a.
           ZZ   ••• O O IV — 9> O •* •« ft»« >» « » O Al O
           oo   o o«v o «« OIM «• r» in a •a o 4 o o « o
           «• O   OOOO0*-«^»*l<*OOOOOOOOO
           ». r   oo oooo-»«t« oo — ooo ooo

           oo   oooooooooooooooooo
           X Z
            zo   o-
            O2   •*«•
            -••*   ooM«wn*«tfi <*n o oo ooo oo o
            » >.   o O O « M O MO O OO OO O O O O O
           o •*    a o o
           a. z
           zo   o » ft »«•*» n* o
           O Ik   *•<• M O ^ O O O, M ^ ^ A ^ O O O »» —
           •».•   O — «- •• rf» O O *» Irt «•— O O O a a O »
           i- Z   o o oooo o o o ooo o e a oo o
           X M    ....»..•...»«»..»•
           a. M


           20   (««««•«•#• a«l«IO«OMM •*«•<«
           oa   o o o o «- — » o « » — 1« o •« ft*» " »
           »^   o o o o o out-* A « oo oon a — o
           ^x   oo o o o o o o « «-«o a o o o o e
           OCtal    •••••••••«••••••••
           O O   OOQOOOOOOOOOOOQQQO
           a, •«
              at
            O «   O O a M Of»
            — at   oooeo-«oooeooi»-o-«0'~CT
            >_ «   aa a a oooo oooa« «»«°OM a
            a z    •  ••••«••••••••••••
           a. I

01
           o«    o o «•  o««of* o*»
1-1            u    •                                                O«   a o o a a o a a a o o a «\ —««t <•»»
                                                                   — O   o o o o o o o a a o o o«» o o *» «n M
           Z»    « M « 4 « ..«•.».•» ok o •• •» to -*M a «    ^-—   a o o o a o o a e a o o o a a •* « o
           — a    e a o o o *«-» o->»• e ••»•»*>* o    O3   o « o a o a o o a o e a e o oo o i
           *-«    a o o o oo — a e M«»OOO o-*»o    a, I
           arz     ••••••••••••••••••
           o<    oaaoaooooooooooooo


                                                                   — *>M O•                                                           ••A,   O O O O O O O O O O O O O —•««-= w<«
      ^   z>    M — « »- o on •* o *>o *» * ««rf»««^> a    ^ z   o o o o oo o e o o o o o o o*» « o
      ac   a«    ooooo»«-o««>n»f-.— .^oo — ^    QE.U«    •<>«>••••••«••>>••••«
      o   —at    ooooo»oooo«ncn   oooooooaeeoooooooo
      a.   »4    o o o a o a oo o a o-~Aitf» ooo o    a.
      o   zz     ««.«»•«•»»«•«««•«.
      x   o«    o o o o o o oo o ooo oo ooo a
      a,   a. «                                           >         7?   -M«tA9O«-O
      u   Z a.    o o o o oo o— ••••••OMM o«-n» *    »> x   oo o o oe o o oa a »• o o o o o o
      Z   OM    ooeooooaooooo— *»~f=n*    aco    »««••«.  »...««c.«c.
      <   — o    o o o o o o oo o o aoo o oo » »    o ac    OOOOOOOOOOOOOOOOM«    a,o
      oe   «ui     •«.»».•«•»»••«•««•      M
      «   O»    aoeaoooeaeeooaeoee
      >   a, Z                                                    ZZ   »•»•» o «i ix o -•
                                                                   — Q   a •« <• M o o o a — o o o o o oo o o
                                                                   »w»   onooooooooccooeooo
                                                                   Ota    «»««c«»e««»«»»«.*»
           z «    e*>»e««««nia>im«»    c ••   eooaooeooeaeeaoooo
           o tut    o — .« *• «n -»«i— o -»-o i/i .a g» wim « o    a, u
           ~ o    o«M»-*n«tmoiM»«      z
           — > z     ••••••••••••••••••
      M   •» —    •» •* «i * ^ »• « » — M n*« (V *« o« At «    z o   e •» « iw« »»i"i o> <« •• e »» ft «
      o   o                          ••••••
      -•   z
      —   o                                                      *= a   o o on«m a a -» a a a o e e o o a o
      UIU                                                      aE««e«»e»»«»e«»«e»»e.«
      o                                                           O      e o o o o o a e a e o o a o oo o o
      Z                                                           a
                                                                                                aaaooeaa»
      ••      o      ooooaeeaaeoooeooe   o      eooeoeaeooaoaaa
      OE      —                                               a   a.
      J      I                                                      I
      -IT                                                      T
      03                                                      3
      oz                                                      I

-------
                                                                             4-103
 o
U
 o

 o
      OKI



      O      o O OO O O OO O O O O O<




      Ou*   a o o o o o on •« •* o a z



      o z   «• ««o«« •

      »->-   o e o<»i
      oc ix    •  • •  •
      O «   o a o e 4



      Ze   ••»• a a «•* a a oiwtft « « a -.-.



      K K    •
      Ota*   a.
      & H»




      ••••   a oo oo <* ontf « •• o« o «»o
      •-»   a oo o o o o a — * o aoo o o
      XI*    ••••••••••••»•••
      ou   aoooooooaooooooo
      a «



      o «   a o •« »<->o«i « « r»«•»«•« «i


      X Z
      O •«   '
      a, M




      — m   o o oo oo«i oo«oi» — « ao
      » a   o a a o o o a a a o o o»» — •



I    II   ilSllSSlSsSSioaS             If    o»ooa*e'ooo'a'ea'a*a',
OL    »-•<   a a a a a a a a a a — *-a o a e
o    at r    »»•••»•••••••••»



tot

2    oiu   ooooaaaaaoaa»»Of-«
«    —o   s» o o o a o o o a oo o o a -a ni             »
—    »x   ooaaaoaooosoacM*.             *
X    OCtal    ................
«    O*»   aaaaaeaeeaaaeaaa             22    ao — — —.-?—«o< — » — -.i
      *                                                          O *V    O O O O C *• O C  9 A0 O f^ ^** O '
        ^^                                                       ** tti    9  f   9 9 999 O 3 O  O O f*5 43 •"* 9 '
^>    O                                                          * ^J    •*••*•••  ••••••
t»    o                                                          oueaaooaoetoaaaoo
?                                                               ».a







^"      **•      •••••••••. ••••a             o*»   oaaaaaeaoaaaaaaa
—      e     a e a a a o e a a e a a o e e             S c*   «"*•«» >» w « w o  » a a w a o a a
X      ••                                        32

Ui

*•      (^                         «««««.««^^^               ^                          ^ ^ B^ •« ^ «^ ^M
-I      I                                                         »
-IT                                                         T
O      3                                                         3
^^      ?                                                         3f

-------
                                                                                4-104
             Z>—«  OIOO^'O-O'O—•« 0-0.0-0«»*"O'O-«
             a 4   O -O O O O O O O O O -O O O O O O !•» tf%
                                      oooooooooo*
                                         «»» •«<• oi» — « am
            OO-   O»OMO>I» O-0» ««,•«•»»•-•.» O-O-AIO
            »U   O OO 9 » O OM*> O O •* OMO O O O
            — «   o oo-o-O'O-o * an* a--» a o 0^3 o o
             * o
             az
             2O   O «
             OB.   — owtMtM* —• • •» 
             Q«   o oo o -«o o o •*•• ••« e o
             ~—   o oo o o o *• a — « m —
             ao   o o-* •> o v • o *»*»'           _
             -•«   OOO»OOOOO«*OOO«P»OIMO
             *»Z   O 0000 OO O 000 O«*...»0 00
             OCM    >«•>••......»•.»•••
             aZ   OOOOOOOOOOOOOOOOOO
             0.
"*           Z-   I*.
             a o   o <
 ..
 ft
 **
•i-l
 C
 O           3u   o ooo
            Q O   OOOOO«M«
            «•<   O 000 O O O O •» -= O 9 Ht ««t » « O   Q3
            ^ Z   oaooooooooo«»oi**v»»ao   A. X
             O •»   oo
                         »900*>«Oni&O«>o|p«>4«   KOC
                         ooooooo«»nir»»oOMi«io   O A.
                         oooooooom»oom«oo   a.
       z   OL.                                                       x> x    o o — o o-»s» -• o
       O                                                            O«ooo^^oooo
                                                                     «- a    ooooooooooooooo^vno
       C   ao    OOVtf»OI0%«O9IVO^9lll^««-«O   OtlM     ••••••^••••••••••)«
       O   •->•<    O OO — O O »»> O «•»» » —•« « O — • O   Q»    OOOOOOOOOOOOOOOOOO
       X   ^ Z    OOOOOO»OOOIVOMOOOOO   A
       O   KM     •••••••«•••••«••••
       OC   OO    OOOOOOOOOOOOOOOOOO
       •k   OL                                                       Z*    mo «.0 •« O O » — < « in«t •>•»»«•«
                                                                     CO    OOOO*9090VOO«M««.0»«
       IM                                                            »U    O O O O O O O O O O Wt •« O V* « 0    a, z
                                                                     —> Q    OI/l<^O«e«MOOOOOOOOOOOO
                                                                     •>•>    OtBOOOOOOOOOOOOOOOO
                                                                     ««     .««••««•»•»«•«••««
                                                     a« aaM   Q«    oooooooooooooooooo
            O LU   oa«>no«»irio»>v->«o»in»-*   &u
            ^ Z     ••••••>•*•«••»•  •••
       •   ««•    •»«K«Mt4««OO<«t9tft«3«—>9   ZO
       u   o                          M«««i«.MM
       *•   a
       •»   «J                                                       at
       9                                                            <=
       z                                                            a.
       O      «     •o»Mo««wt*«o«f>»ma>n>«    C •>   ooooooo — OOIMOO~«IO»«»
               »    •r>«i»»'«»«M«««tf««i»ooni    •=>-   oooooooo — ooooooe««»
       »•      Z    —*•*»«•«••«,.-. -. — ooooo  o o a    »•<«   oooooooooooooooooo

       -•      13       OOOOOOOOOOOOOOOOO    O      OOOOOOOOOOOOOOOOOO
       Ot      «.                                                O    Ok
       «      w                                                  .    "*
       w
       z      at    —
       M      U
       J      S

       o      a
       ^^      Z

-------
                                                                              4-105
      3 «    9
      -»*l   a.
      —       a

      3       999990aCiaaa93aaa 9-9 O
      S. f    «-«»
      33.   aa«'
      -. -j    ooo
      — X    O 0-9 0'9 9 9>»  O—9 — O-«'9-9-9 9 9.
      X 3     •  ••••••••••••••••••••
      3 U»>   9 9 3 9 9 OO «  «^ O 9 9 9-9-9S* 9 9-
      — >•    OOO9>««9-»99999999O<
      xat     •••»••••••••••••••«
      3«    ooaaaoooooaooaaoai
                              o »  •• •• o »w •* a -• m »•»
                           n
      — I .•    onioa«OOO-»OOOOOOOO
      — I    O, a * •• ••OO a>  >aO O) a 9, a a 9 9
      x ««     .»•••«.•«•••«•••»•••
      3W    O 9 9 9 9 O 9 9  9 9 9 9 9 9 a O •»
      at al
      OU
      tL <
        Z
      39    9 9
      «. ^    O 9 9 — 9 O 9 O 9—9 9 • » »• 9 — — 9
      — f    9 09 9 9 00 O 9 9 9 9 It* — — 9 3 9 O
      «ni     •••••••••••»«••••••
      O Jt    99oaoaaoo>oo9oaa9999
                                                                    O <    9090999a99
                                                                    •• O    O999999393
      r—    4 9« ««i 4 » a m o» i«f« « 9 -tU   99909090999990149990
                                                 <»— »    ac*    ••••«.•...«........
                                                 -o«*««* — o    O a.   9999900000999999999
      O •»    O009009999999900OOO
      &                                                             ZZ
                                                                    94
                                                                    •- O.
      z—    ni — o » m «» 9 «^w> » « MM* a — •«    O  O O O O O O O O — — W • 9 O •>* — — O    O •-    a O 9 9 99 9 999 9 9 09 9 9 9 99
      »Z    9  99 9 9 00 O OOM — O 9M •» 9 O 9    X
      ati     •••••••••••••••••••
0    O •»    0999900099999990990
z    a,
O
—
»-    7->
oc    30    ooo<».omtno»«»-ooar»«« — of»o    Ota
3    •• <    OOO-»OO«^O«W^»»I>J«»O — « — O    OU    9OOOOOOOOOO9OO9OOOO
&    »Z    O9O99O«9OO<«O*tOOO9OO    A, O
O    9M     •••••••••••••••••••     Ut
Z    O O    9
                                                                    ZZ«M.a.wQ<*
                                                                    O >•    OIW»9OO
tot
W    23.    99999>>O9OOMn<9«ni->m«n<    ^ •»    9rbo»9OO99999999999O9
Z    3lU    O99O9O99OOOO9O4O9*^    Z Cl     ••>••••••••••.••»..
«    «u    99O999OO9999OO9O-AOIM    O •»    9OOOOOO999OOOO99999
-,    h.x    oooo909eaoa9O9O9onir>    ft. u
z    or jj     •«•»«••••«•»»«••«••       z
4    O "-    9O9999OOO^OOOOOOOO9
>    a, r
                                                                              9999900999000900»0
      2 T*    OW*« — «*«-O'«»^'"-»-»    Q      O
      ClU
      ~ Q
n    «••    —t
O    O                            «.«•«.••• (V« < « »    OU    0000000000090000O        •                                                    ^      e9O9OO9OO99999999ir<9
«u                                                             z       •••«•«••••••«««»•».
O                                                                  O      OOOOO9QCO99OOOOOCOO
Z                                                                  ».
u      u    9««~«<«inem9M«v««-«m«f»^
«      3    o»t.»n»».«-«ni9 010 m<1'
                                                                    O •»    OOOOOOO— O
                                                              '>>    •»*•    OOOOOOOO —
>•      Z    •»^lfttf«4it«*«»«>*»O9OOOOOO7    »•«    OOOOOO9OO9999OOO99»
*•      UJ        •••••••••••••...  .9    Bfc     ...................
~      (J       000000000090099^000    O      9OOOOOOOOO90O9O99OO
*      ••                                                    9    &

-------
                                                                                   4-106
          O •*   o o o oooo ooooooooo <
          Mtk.   000000000000*0004
          •-      o oooooo 00000*000*
          at       ••••••••••••••_•«
          a. ••
          *• X    O 00000 OOOAIO—«
          aco     ••••••••••••
          ou    ooooooo^oo«o«
          ». >•   o o OOOOOA««<»O« o» ooooo oo o
          ac ac    •••••••••••••••••••••
          o •«   oooooooooooo^ooooooo-o
                                                        000000
Olt    O MO •»»*«>••*»« « *«^ •»
....    o OM — mo o«too o»-*oo
»- X.    O O 00— WIO 000 O 00 00
                                                                          OQC   O OOOO 00 0 — 000 OAtM<» *-l
                                                                          ««O   o ooooooooooo oo —ntt*<
                                                                          »••   O OO OO OOO OOO O O OO OMO O OO
                                                                                  oo o o o oooo
                                                                                                                 «••«•* — AIOOO O
           Z **   *fr IU ^* •** ^k ^fr ^^ m MK MK p^ tfuf^ O ^f ** *• •*• ^B ^9 ^B   ^* ^rf   fi^ "O Cft v* <> ^> 4M |W •* 4k 4^ ^> ^t ^9 Cl ^ft 4V ^^ ^yt ^^ ^%
           OB   O OO ••«•«« O OW«0 O»»* •*• O •»»««»-*»   OL K    ••«».•««••••«.««»••»•
           O ^^
          *• at   o o
          OC9    ».
          o •»   o o
                                                                          Xt»
_.                                  %                                      ••«•   OOOOO'
3         Z«*   MOO«MI—• <» ai» •«» OAI»-»»— «KA**«   »-O   00000«
           oo   o o —«-»« o n»«
<         —at   o o o o o o o ont««% ot^-»ni   ac _*    ««..»»••«••••«•»•«»
           -»-«   o ooooo o ooo o o o oooo** —MO   oae   ooo ooooooooo o oo oooo«
           t-T   o ooo oo oooo o o o ooo a o«»wo   *, X
           at >    ••••«•«•••«<••»••••••      o
     •S    O •«   000000000000000000000
      o      o
      OC    OO   O^OMOOOOOOOOOOOMO9««W«»    « C    •  •
      o    MO:   o o o o o o o oo o o o o o oniotft o »»o    ou   oo<
      a    ^ *   o o o o oo o oo o oo oo o oo o««« o    ^o
      O    OC*«    «»••«•••««•««•»••••«•      UJ
      flc    o o   o o oooooo o o o o o o o o o o o o o
      o»    *,                                                               zz   «•> •
                                                                           O >•   O«4<
     O    Z O.   O O 00 00 O 0»«0 O O 0«*OOOO OxO    1-0   OO
     i    OKI   ooooeooooooooo«o«*»o»M    ff O     • •
                                                 O O O^ O O O ••••    kU
      <    O >-   O O 00 O OO OOO O O O OOO OOOOO

                                                                                                         o OM o«»o OM*«> o<
                                                                           fr>O    OOOCOO^OOOOOOOOOOO««O

           zx   o — «^ «» o nj>Min « r*t~ e —1*.» » « o*»«    o      o o ooo oooo oo oo o o o oooo o
           O 0« O o >eVO AI«MNtO   ZM    OM O M O O O O «t ••
      «-»   9                                 ••«•«•««««»>•»   ot*    o ooo ooooooooooo oo« « o«
      *"5                                                         ••   *• «    o o o o o oooo o oo oo o oopt » ««o
      »•   o                                                              ^      o o o oooo o o ooo o o o o oo om o
      o
      2
      (9      iu
                                            o •« a » in « o»» ^ -
                                            m « •• o —iv m » m •« t*.« » o«»     oc    —«y« »
      ••      >U                          ••••••>•••.•.•••.•«• A| Ml     ul
      J      05                                                              31
      -J      ».                                                              X
      03                                                              3
      «      *                                                              Z

-------
                                                                                   4-107
                                               ._  _
           .. -.   00— IMOOIVOOOOOOOOOOOOOOOO

           3 ^.   ooooooooooooooooooooooo
              u

              Z   »•»•>»» — O O <
                   009 —O O —«
           ^ JC
             .
              M                                                                  O«    00 — to*  o — e a »»M •» o — *. •• » « « a •* «•«*

            fa    * ««|fcf» » I* » » * «t • •» «•• » !»*»••»-•»» W «t»    »»tJ    O O O O S>  O O O O » O — O O O O O — O O O O O
            ..—    oo * MOW ••••* «•>«» onflio nt»o o oowo    oo    ooooooo
            »•»    OOOOOOOOMO— oooooooooo — o    *. z
            at «     .......................      -*
            O ^    OOOOOAOOOOOOOOOOOOOOOOO
                                                             '
                                                                                  O*    O O O O O O O O O 0«- O O O ~ O O »" W
                                                                                  MO    ooooooooooocooooo««i«t»w«"O
                                                                            —    ^*»    ooo o o o o o o oo o o-o o o o o e ixt •» o o
                                                — o» o«i o ««o a o o *t    aa    o o o o o o o o e a o o e o o o a  e o e e o o
            *• f    o o o ei o«»«io «—o o o o o o ooo oo o«—
            XX     •••••••••••••••••••••••
                    o o «o o o o o o oo o o o o o »c o o oo o
                                                                                  za.    »» ••• o •»•* o •» n —
                                                                                  O«-»    OO«» • M « O OM*» « Ht Ml» O« VO> n( 9 9 O O
                                                                                  ••U    o o o »• w »» o —«•!•»•— — o « e «••* o «• e o o
                                                                            o    a«.   o oo o o o o o o o o o o o
                      o o e o o oo o oo — o oo — — m o o o o o    a. »
            Z ->   V »>^ « M ••*«<«•>«« 0« •»^«*l>m« «0*    »-M    OOOO O O O O O OO O 00 O O O —•* e —O O
            oo   ooo«»»»oi»f*o*Ao«A><»f»««»t»««o-*    ae«    .«••»«»««.«»».».....»..
            »a    oo o o o o ooo o — o m —• •»«* ~»in oo o    OX    oo o o o o o o o oo o o o o o e o o o o o o
            >_ ar    oooo o o ooooo oo o *-MOO oo ooo    «.>•
            ac
               x    o oo « o o o oo oo o o o o o o o o o o o o
                                                                                                                                o « — — »>.
            MO?    o ootloo o5«-«t* o—«••» o^oooooo    oo
            v- z    o o o o o o o i* o oo o o o o «• o o oo o e o    ^K
            arc     •«••»••••••»•••••••••«•      u
*2

            Z—   *» « ««»»o • •«*§>• ««•«•» ^» Mm —o* «»•—-•    ^-X    o • o o o o o o o oo o o c o o o o o o o oo
<          Oo   ooooo^^«»o«ow»^«»o —oo-«ooo    or j   •••»••>••••*•••••••••••••
            — or   o o o oo o o«-o oo o«» « o»> mo woo o    OK    ooooooooooooooooooooooo
                                                                                    u

                                                                                  zz
                                                                                  ao
            ao   o o •• o — o — o •>«»
            MOC   O O O O"» O O OO OO O • <• * * •"«•« O » O O    QO
                                                                            o
                                                                                  *o    •« o o «» v ~ •» «ntm*«m « r> o »•-••« ^ M««
                                                                                  az    o^»m»» •• » « o o»» « P« — o o to o •»««> v » o
                                                                                                    o«roooooooooe>oooo~»»o
                                                                                                    o «»o o o o o o oo o o o o o o o o o
       K    Oo
       a    <->ar   o o o o • o o «•>«*•« o « o -•«••« «*o mo o    oo    oooeooooooooooooeoeoooo
       a.    >. r   o o o o o o o e o c— o o o o o o o »»o o o o    0.1
       c    ««••   ••«•••»••  •  •  •••••«•••*••••      «

       a.    a.                                                                    za    OM»»O •*»•««> •». M« o»»»» — ...—  90 •«*«•

       hi                                                                         "»«i    ooooooooooeooooeoooo«mo
       u    z a.   o o o o o o o o oo— o o o o oo»» — «•»•    ^- o oe oeooeooooooo
                                                                                  ZM    oo«ow«v»oMeM — «>«>oeoK« — i»«eo«>«o
                                                                                  ^      oooooooooooooeoeooooe»o
                                                                                  O       OOOOOOOOO9OOOOOCOOOOOOO

             — o    o «M — •f^o »»*•<» «o4«»«eA>o«0»
                          _ _      ' tf^ — ^^ ^*o ^ o •• «• v* At pk> «• 4 ^n i^ «•    z 9    o o o o ^^ o *• o o o v^ ^* ^^ (^ l^ At ^* •• o o ^ flf o
        u    o                                    •• •* — ^ ^ ^ M m^^o    OOT    ooooooooooooo^ovovo^p^^^
        *•    Z                                                             M    ••»-    oooeooeoeoeroooec'Oeeo — o«*
        ^    O                                                                   >>4oooooooeooooooooeoooooo
        •>    u                                                                   z *.     »»••€•»•«••«••«»««...«••
        O                                                                         O       oooooooooeeoooooeeeoeoo

        e      M    MMMO-COVOI	
                     —•»«« — » v< o«ir>wi •• 4* mm
                                                                                  o«    o o o o o o o o o o o o o o o e o  oo o 9 « 
-------
      oo   e o «>«*« m «•» «••«» ««M«VI o««>
                             -    ..........••
     O k>   e e o o o o o o oo e o
        u
      ZX
      oa   o « o •«•«••*•««•>••• M o MO«» •«
      — u   o e e e a e o a •»«•» —1» « e «•* ••<
      at o    •  •••••••••••••••••••••
      cu   e o e o o o o o o o oo o ooo o o o o o o
      a. z
        to*

      Ze   o« « o»* «o«w* o« «.«•#*. g» ««• o«"» « c
      ••«•   o a e iw«v ••« otn o M gi on e -• o o e o« o
      »- >•   oceacoooPtMooooooooooxo
      a «    •  .••«•••••••••••••••••
      O <•   e o o o o o o o o o oo o o o o o o o o o o
      tt. x.                                                              Z*.   -«o o o o OMO o-* o*t««
                                                                        MO   o o o o oo o o o o o o oooo o ••«  o •« •»«•» « >**««• oa> •» i»  o oni«» * « IP  o o o rn«v« tv o «•» o o o
      ZO   •» «».-^*fc •« »O •«»>••• « tf»OOM «•••' •••» <»   *•*»   O 00 O O 04W-00 IM O O O OO O O O O O O O
      C >»   eaon»o«io(wo>»Kl«'«ltWf^l>n^ni«l«>M   arc    «••«•»«».«•»»••»•«•.«.
      ••••   o o o o o e ooo — «•— •»p» ««o»» o«f •• o o   oa    ooooooooooooooooooooea
      t-f   e o e o o o o o o ooe o O

        X                                                              O^oooacoaooiwoooa«o«'»m-«m«
                                                                        ••*-    aooooooooooooooa<*»«e«'—c>
      r —   •»«•••»— — en»O'»»» — «»^O9'»»»««i»o^   »-»    ooooaooocoooooco—«e«°cc

      •»*   o o o o o o o — o o « e — « -*nm •••• o o o   OT    o oo o o o o o o o o o ooo o oo o e o o
      »- Z   O O O 00 O O O OOO O OM«I OOOO O O O   O..>

                       .000000000000000000

                                                                        OZ    0^»OMmi«««»>

                                                                        »-o    oooooooo
      oe    eaoo^«a>ooop»
      — «    oooo^oon*^«-*ooa»«moooooo   oo    ooooooooo^oo
      N> ?    o o o o s o o « OM o o o — o e o o o o o o   &ae
      aro    .»••••••••••••••••••••      u
      o o   ooooeooooooeoeooooooao
      &                                                                 ZZ    «o«k •«»«•».<<»••» OM
                                                                                                 •• *l »l  o « o o oo o o o o  o o o
      Z»»   jnioo**^**^*-"****!*-*****"*!^   *•*    o«oooo«*ooooooooooooooo
      O o    cooo«i — ««o-»n« — u««»»o>oo««oao'aj     •••••••••••••»•«••«...
      —«    o c o o o « o o o»» o ••.•-•—» « e «o o o   oar    o o a o o o oo a o o o o o o e a e e  c- o e
      •- y   eoooooooooo4to-«««ooooooo   ft  X
      ar«    ••••••.•••.«««•••••••«      u
      o«»   o o o o o o ooo o oo oooooooo oo
      a                                                         •        22
      f -«   «» — • ^ «i -•*». o« •» o o «*»<»<—«» o   fc. T    o o o a o o o o o o o o o o a o ni «w 1*1 o o o
      oo   eon«omoe*oa««>aM'Oi«M-*onooMo««*«iooonoQ   oo    ooooooooooeeoooooooooe
Ik    •- o   ootftoooooooooooaaaoooao   & X
o    « «    .»«•••»»••«•••••••••«•      M
!T    O7   OOOOOOOOO9OOOOOOOOOOOO
a.    *. ri                                                              ZO    out o« o «*M« — wooooooeoocooooooooo«co
u    Z*   o o o o oo oo o-«a o o oo o»- •••» o •* o   •-«    oo o o o e oo o oo o o oo o o o o e- » o
z    o M   ooooooooo9oooooooa«««......
•«    ~t»   o o o «o ooooo oo o o o o oo omt* •   a      o ooo o ooo o o o ooao o oo o o c c
•«    fc. at   o e e o o o o o o o o e o o o o e» o o — o*-   a
•x    «• w    ««»«»•«•««•••••«•••«••
«    O ^   eeeeoooooooooooeeooooo
                                                                                                                                   of<»
                                                                        «-«<    ooooooooooooooooooo««o
                                                                        »-      OOOQOOOOO<9OOOOOOOOOQO>0
                                                                        ac       •••••••••••.•*••«.«...
                                                                •»»    o      o e o o o o o o e o o o o o e o o o e o e a
      O M   «» O «••«••*'•»»-•«»« «»«1«>  OO •»« ^ »•»    A
 •»   •- ••   M««««t«l«««»»**OO«»«^A»«ai
 O   a                                    •».«••«•••«• «» 

 o     M   *.— ««.•»«••*«•« o « ««t «•"••»•«»»» *e«
 «     3   « «» m •»• « o o<* IT *«••> •* v -• ftt e o •« « « -»
O      «    ••>»  o o> o « •» o «••»»* e iv^. IM «• « «•» i   tf> •  »«» «*» O« • ••«» •#««» —fc •» WOO -•    —*.    O O 00 O O OO O O O O O O O O O O -30 OO
>•      z   .»o>  « .0 «t •«•«« m At M-> -»•»•••»» o o o o e «    *>      o o o o o o o o o o o o o o o o e e o o o v>
•_      M       «..«»»«••«••««•••«••<»    «      »••*««*•«•«««.>««,......
•-      o      ooaooeoooooooooaooooo    a      oooeooooooooooooo ooooo
«      —                                                         o    a,
7*
                                                                                                                                o — iv
 _

 S1      §                                                                §
 O      Z                                                                Z

-------
                                                                                4-109
           oo   o o • n M n « o ^ — w < « o ««i Mn MM o o

           ^. —   c o o M a — o o o o — o o o o o o o o o o o

           o »•   oo oo o o oo oo o o o ooooo oo o o

             u

           ZZ   O— — *l — •*O^^^O*^OMI
           OO   OtftOo   A! :
pa   oooooooooooooooooooooo
 » Xk
                                                                          O ••   O O n».O- n n; ^ • 0k 99 ^ o o ^ f^ • • Ml ^ O O

                                                                          ••O   O O O -»O « -«»«-*-"«*»-»»«»r»   oc ac

                  oo o o o o oo o o oo o — o *» «> e o o o o   •. >
                                      oooooooooooooo
                                                                          Z*.   •• O O •• O « «x« •!• O A* «l
                                                                          ex   o ooo ooo^ — o o w o M « •
           OO    OOO«—*»MOO-« —M O •
           — or    o o o oo o o — o — o* o.
           a.                                                             Z«   «o«w-«« «

                                                                          ••••   O O OO OO
           oo   ooo*»oo»»"-^» w o««*»n»»o«o«*   at x    ••••••••••••••••
           ~ .««
2                                                               -03   o o««>^ m«i«e o vtMo o o»«o o o o-»o
J                                                                         —O   O O •«» 00 •*>«>« O •• O O O O O O O O O O O
           Z —   AM OM*> •>« M o o»o «• •^OM«IO«I^   ^X   onoooomooooooooooooooo
           •-at    oo oo o o oo MO o» •» ««•• «oi-»o o o   o«   o o o o o o oo o o oo o o a oo a o o o e
           ^Z    ooooaooooo
                                                                          zz
                                                                          oo   o o e> o o o-«o o « «. o o •* o»«>ta> o o o
           z—    «o «•« »» «^.o — < «« ••oo»«»^no   ^JE
           oo    ooo*) — — oo«««i •— • **nn**»a»o«fOM   « o
           «• at    ooooo»«oooo o» o»»«(*>ooo«oo   ou
           K- z    oooooooooo oo ownmoooooo   CkO
      ••   O Ml   O O O O O OOOOO OO O OOOOO OO 00
      Z   ft.                                                             ZO   •—<
      o                                                                  oza —<                          _._..__.._
      —                                                                  ——   OM»OMM«O—00OOOOOOO———O
      i-   z—   « * «n — n •*•«!•>• M«—« o — «» «^» •>«•   ^>>   oon — — aoooooaoooooooooo
      K   oo   ooon«o» —«*> •»«—-• n —n —«>MMO   acx    ••......•..••..•«..•..
      O   —at   o o o o o M on « •« M«IO» mn — • Mn o o   oo   oooooooooooooooooooooo
      a.   N-z   o o o o o o e o o — oo o o o o on oo o o   a. X
      o   an    «..».......»•••••••••.     M
      K   oo   oooooooooooooooooooooo
            .
                                                                          O«t   o>.oo 00*10 MO — o«t —o ooo-« •« am
      M                                                                  — W   O O 00 O O 00 OOOO 00 O O OO O « — O
      v>   z  a.   oo o o o o oo o—>oo oo OO»«*«H« oo    >-o   00000000000000000000*0
      Z   ow   oooooooooo oo oooooo «nnt«    « <    >•.«.•••>..•«..»......
      4   —U   OOOOOOOOOO OO OOOOOOO*t*»«k    O      OOOOOOOOOOOOOOOOOOOOOO
      —   *•  «   oooooooooo oo oooooao — o^    a,
      CC   X  Ui    •••«•«.••••••••••••«••
      ••   o  ^   oooooooooo oo oooooooooo
      >   a.  z                                                           ZM   OM»—o —o« o A»—M —«i-o —n —M «
              —                                                           oe   o o o o o o o o o o o o o o o o o o o •»
                                                                         H-      OOOOOOOOOOOOOOOOOOOOOO

           ZM    o^nn^n -•» » — «o »••>»>. — «oo — oo   o      oooooooooooooooooooooo
      u   o                                  —————MM«•»    OM    oooooooaoooo*—oOMon«M«
      —   Z                                                        —    — *•    o o o o oo o o o o o o o o a o o o o — o «
      •-   O                                                             »• «    O OOO OO O O OO O O O O O O 00 O O 09


      o                                                                  o      oooooooooooooooooooooo
      Z                                                                  Oi
      «      3    — o — «• « *• oon c «• — n « onm « o * —
      —      -J    ««« lO *> »••—•• «» •«— nn — «-».«p. OM O    Z^-    OOOOOOOOO — OO«OO
      O      «    ••«o««»— «i»^ <•» Mont — Mta»p»«Ai«    O «    ooooooooooooooooooc-»-««
              >    ««»-»— o«^-«»» — ». ««tn« — ^i-vAtoo—    — ib    oooooooooaoooaooooooo«
      >•      Z    —»••»*»•• »««•» M— — — — «. o o o o o »    •-      oooooooooooooooooooooo>
      »-      W       ..«••»...•........,  ,o    x      ......................
      —      e      oooooooeo eo oooooooooo    O      ooooooooooc
-------
                                                                                  4-110
            ZO    o ooo-»oo o «n—«i o»t,« o f»« o~
            on    a oo o o oo o oo o» — o 
            Mfc,    o o o o o o o o e a o o Oo o o o o
            at
            o
            Za    o
            oo    •• e »
            •••»    o o 4
            at >•     •
            o>-    o
            Z Z    <• '
            O O    01
            >• X
            or o
            o  •*»• *4 —•*•» I
            *• *•    O O •ft^*^* ••M •     o o o o o oo * MO o o o o o e oo .
                                                                          OB   oe o e o o oo~a oo OM ••**> n M«I «
                                                                          MO   oa o oo o o ooo o a o O«M *»• «»• ««*• O «• •»•»   ^M   O O O O O •» O O-e o O O OO OO O O OO
            O O>    OO-»0«MCO»»MO><-*
            aiu     •»•«»«.«..««•••«••»•      «
ni          OO    O O OO O OO O OO O O O O O O O O OO
            a. •«                         .                 '               Z «.   MOOMOO OM*«* mooooo «§«oo
              at                                                          OS    o o o o o o —-.ci — «» o o ••»— <•»«•»••»
                                                                          •">-   o o o o o o oo o o o o oo «•«• e>»»«-»o
            Oo    aoa-«>«oato

            »- z    o o oo o o o o oo 4 o«=o o o oo oo   fc :>
            ac «    «.««.«....«»«««««»«»      <
            OM    O O OO O OO O OO O OOO O O O O OO
            a.                                                             za
                                                                          OZ   o ««* — e o» ««*•« tv *•«-•«-. o
                                                                          MM   O O O O O O •» — .— *. 0>  91 t\t O O O * (UM« O
            z —    *» o—— «•!•»•• *i «••••.««»«*»«••* *»•*«•»   N- a   oo o o o o o o om o  — o o o o — o o o
            OO    QOO|0.— «»M«I ~>0 o«P-«*a.OO»««W   XX    *•••«..•••««.»..»...
            —  z    o oo o o o o o oo o o •« «woo o o o o   •. ac
            aa.    ••«••««•••••«•••••••      u
      M    o *>    o ooo o eo o oo o o o a o o oo oo
      Z    *                                                             ZX   «•«»—•• — "§••• ««0*»—>-»» «•««•
      a                                                                  OO   o w «* -»o o «•» o •» «  o o— o o — o c» —
      •_                                                                  »• a   oo »• «^ o •»!«« o ft o  o o o o o o o a o
      •-    z—    « ••»• « 0 «<«•(«-••« — • * *• •• ••• «o   fci   o* o o o o ««im««   K.J    •••••••««•«•»«««•«.««
      O    ""C    OOOOOtfl««4Mlkb^nOOO*^9O^«OO   OV
      a.    •- 2    o ooo o oo o MO o o oo on oo oo   », x
      o    an    ».«..«.•«..«•-••••«•      «
      •«    «-t»    o o ooo oo o oo o ooo o o ontr- »   oo   oo o o o o oo o o o  ooo o o o oo o
      -.    >-     AZ                                                          zo   oo •*»tfl«'« *»«wo — ««  o«w«« worn •« «
              — >   .                                                       OZ   -»•••)«« « « o o ••«•  o <* « « o « (w« o
                                                                          »•»«•   o «• * •••» .»fc O OO O  O O O O O »«-«»O
                                                                          k >.   oo«o»oeooooooooooooo
                                                                          IX    .•««»«».»»««»»•»»•»•
            ZM    o«««r. ^•••-»K»«»-O»>.«*»«W«<«»   oo   oooooooooooaooooooo>o
            o la    o »»*!-•««•••«-•-•
            •• O    O O ^ *• ^ ^ • 10
      «i   MM   «• n •*
      u   O                                   «««»•-» <««(«•>•   OW    o Jo o o
      ••Z                                                        ™»i««uo
      ••O                                                             I^OO
      KtO                                                             ««««•»»«««»««««,«.»«.€.
      o                                                                  O      oooooooooooooooooooo
      z
                                                -»-"-»ooooo-*    ••      oeooooooooooooooooa^o
      »-      u       •.«««••«••.««».... o    at       ...................c
      M      C»      OOOOOOOOOOOOOOOOOOO    O      OOOOOOOOOOOOOOOOOOOO
      KM                                                     OX
      •«      IU                                                      •
      M
      z      at   *• nei -f rf» « »* • » o MAIM ««««*>««> o      *    «•)«« •»•»«•**«» o —«•«*•» « K c •> o
      M      UJ                           .«.«_•-»...•.«.....»««      M                           -.«o.»e— ^-.-. _ .... «(
      j      «                                                            a
      -»      X                                                            X
      03                                                            3
      u      z                                                            z

-------
                                   4-111
                          Footnotes to Chapter 4

      See,  for  instance,   d'Arge,  Kask,  Case,  Ben-David,  Eubanks,  and

Anderson [1982].
     "Tor a  good discussion of  possible non-linearities in  air  pollution
dose-response estimation, see Graves and Krumm  [1981] and Krumm and Graves

[19831.
     "Tor  an  interesting  application  of  logit  analysis  to  asthmatic

morbidity, see Whittemore and Korn [1980].
     4
      Some epidemiologic investigations  of ozone  have att< mpted to control

for pollen concentrations.   See Zagraniski  et al.  [1979].
      See Graves and Krumm  [1981] for  a  discussion  of interactive effects.
See also Bates  et  al.  [1973] for an analysis  of  ozone  interactive effects
with other pollutants .

      This is  the  purpose  of  many of  the studies cited in  Environmental

Protection Agency [1978] and National Academy of Sciences [1977].

     7See Lave and Seskin [1977], PP. 44-50.
     7a_
           example, see McDonnell et al. [1983].
     Q
      Using an earlier version of the Health Interview Survey, Ostro [1984]

also found  such a  relationship  between particulate matter  and  restricted

activity days.
     q
      See Friedman [1981], for instance.

      Others  investigating  ozone-related   morbidity   have   examined  for

differing effects  depending on  time-of -year .   See Kagawa  et  al.  [1975,

1976], for instance.

-------
                               4-112
12Ir
      Studies examining  the  effects of ozone  on individuals with  chronic



respiratory  disease  are summarized  in  Environmental  Protection  Agency



[1978], esp. pp. 208-212.



      Cn clinical  investigations  of ozone  health effects, lagged  effects



are  of  interest in  part because  they bear on the  question of  possible



adaptation to ozone.   See,  for instance,  Hackney et al.  [19771.   See also



SesteLn [19791, PP.  281-283,  and Seskin and Kawin [1981].



    13See Theil [1971], PP.  607-615.


    13a
       See  U.S. Environmental  Protection  Agency [1978],  especially  pp.



249-250.


    14
      A more sophisticated approach to ascertaining the  effects  of  medical



care  on health  may  be  found in  Crocker £t_  al. [1979]  and Gerking  and



Schulze [1981].  See also Chappie and Lave [1982].



      See Goldsmith and Landaw [1968].



      For a recent example, see d'Arge, Kask,  Case, Ben-David, Eubanks  and



Anderson [1982].


    17
      They  included Gilbert  Omenn, Jane  Koenig, Michael  Morgan,  Robert



Frank, Alice Whittemore, and Michael Lebowitz.   John Spengler  also provided



valuable advice.



    I83chmidt [1976], pp. HO, 48.


    19
      See Belsley,  Kuh, and Welsch [1980].



      The following  discussion draws heavily  from  the pioneering  work  of



Belsley, Kuh, and Welsch [1980].



      The work of Mueller [1982]  influences the following discussion.

-------
                                   4-113
                          References to Chapter 4
Bates, David, S. Field, M. Hazucha, C. Parent, and F. Silverman, "Pulmonary
     Function  in Man  After  Short-term Exposure  to  Ozone,"   Archives  of_
     Environmental Health, vol. 27 (1973),  pp. 183-188.

Belsley, D.A.,  E. Kuh,  and R. Welsch,  Regression Diagnostics  (New  York:
     Academic Press), 1982.

Chappie, Mike  and Lester  Lave,  "The Health  Effects  of Air Pollution:   A
     Re-analysis," Journal of Urban Economics, vol. 12 (1982),  pp. 346-376.

Crocker,  Thomas,  William  Schulze,  and   Shaul   Ben-David,   "Methods  of
     Assessing Air  Pollution Control Benefits:  Vol.  1-Experiments  in the
     Economics  of Air  Pollution  Epidemiology,"  Environmental  Protection
     Agency report no. EPA-600/5-79-001a, February 1979.

d'Arge, Ralph, Susak Kask,  James Case,  Shaul  Ben-David,  Larry  Eubanks, and
     Curtis Anderson, "Air Pollution and Disease:   An Evaluation of the NAS
     Twins," a  report  to  the Environmental  Protection Agency,  Office  of
     Research  and  Development,  under  contract  no.  CR-808893-01  to  the
     Resource  and   Environmental   Economics  Laboratory,  University  of
     Wyoming, 1982.

Friedman,   Robert,   Sensitive  Populations   and   Environmental  Standards
     (Washington, D.C.:  The Conservation Foundation), 1981.

Gerking, Shelby and William Schulze, "What  Do We Know About the Benefits of
     Reduced Mortality from Pollution Control?"   American  Economic Review,
     vol. 71 (1981), pp. 228-234.

Goldsmith,  John  and  Stephen Landaw, "Carbon Monoxide and Human Health,"
     Science7, vol. 162 (1968), pp.  1352-1359.

Graves, Philip and Ronald  Krumm, Health and Air Quality (Washington,  D.C.:
     American Enterprise Institute), 1981.

Hackney, J., C.  Collier, W.  Linn,  and  S. Mohler,  "Adaptation  to Short-term
     Respiratory  Effects  of Ozone  in Men  Exposed Repeatedly,"  Journal  of
     Applied Physiology, vol. 43 (1977), pp. 82-85.

Kagawa, J.  and T.  Toyama, "Photochemical   Air Pollution:   Its  Effects  on
     Respiratory  Function  of  Elementary   School  Children,"  Archives  of_
     Environmental Health, vol. 30 (1975),  pp. 117-122.

	, 	and M. Nakaza, "Pulmonary Function Tests in Children Exposed
     to  Air Pollution,"  in  A.J.  Finkel  and W.C.  Duel  (eds.),  Clinical
     Implications  of_  Air  Pollution  Research (Acton,  Mass.:    Publishing
     Sciences Group), 1976, pp. 305-320.

-------
                                   U-11U
Krumm,  Ronald  and  Philip  Graves,   "Morbidity   and   Pollution:     Model
     Specification Analysis for  Time-series Data  on Hospital Admissions,"
     Journal of Environmental Economics and Management,  vol.9(1983), pp •
     311-327.

McDonnell, W.F., D.H. Horstman, M.J. Hazucha, E.  Seal,  Jr.,  E.D. Haak, and
     S.  Salaam,  "Pulmonary  Effects of  Ozone  Exposure During  Exercise,"
     Journal of Applied Physiology, vol.  5*1 (1983) pp.  13^5-52.

Mueller, Charles F., The Economics  of  Labor Migration  (New York:  Academic
     Press, 1982).

Ostro, Bart,  "The  Effects of  Air  Pollution  on Work Loss  arid Morbidity,"
     Journal   of_  Environmental    Economics   and  Management,   vol.   10
     (forthcoming 198U).

Schmidt, Peter, Econometrics  (New York?  Marcek Dellker), 1976.

Seskin,  Eugene,  "An  Analysis  of  Some Short-Term Health  Effects   of  Air
     Pollution in the Washington D.C.  Metropolitan Area,"  Journal  of Urban
     Economics, vol. 6 (1979),  pp. 275-291.

	 and Naomi Kawin,  "Air  Pollution and Health in  Los  Angeles," mimeo,
     December 1981.

Theil, Henri, Principles of Econometrics  (New lorkJ  Wiley), 1971.

Whittemore, Alice  and  Edward  Korn, "Asthama  and  Air Pollution  in  the Los
     Angeles Area," American Journal of Public  Health,  vol.  70 (1980), pp.
     687-696.

Zagraniski,  R.,  B.P.   Leadered,  and  J.A.  Stolwuk,   "Ambient  Sulfates,
     Photochemical  Oxidants,  and Acute Health Effects:   An Epidemiological
     Study," Environmental Research, vol.  19 (1979), pp. 306-320.

-------
                                 Chapter 5








        ALTERNATIVE OZONE CONCENTRATIONS AND CHANGES IN MORBIDITY:




                          ILLUSTRATIVE ESTIMATES








     In the  introduction  to this report, we pointed  out  that the original




purpose of our  project was to estimate where  possible the dollar value of




the  changes  in   health   associated   with  alternative  ozone  standards.




Although  our focus  is on epidemiological  findings  in  this report,  the




statistical  results  from  the previous  chapter  can  be  used to  estimate




changes in morbidity that might result from different ozone concentrations.




And- these in turn are essential to economic benefit estimation.




     This chapter  illustrates how that might be done.  In the next section




we make some illustrative  estimates of the "simple" reductions (increases)




in acute  and chronic morbidity that might result  from reduced (increased)




ambient ozone concentrations.  These are  termed simple because they do not




take  into account  the  possible  feedback  into  acute morbidity  of  any




ozone-related chronic  illness.   Those effects  are  addressed  separately in




Section  5.2.    In  Section  5.3 we   attempt   to  put  these  illustrative




calculations in some perspective by comparing  the ozone-related  changes in



morbidity predicted  from  our  models  with those  predicted to result  from




changes in other  independent variables including  smoking  status,  age,  and




income.  Finally,  in Section 5.4 we suggest  how these predicted changes in

-------
                                    5-2






morbidity  might  begin  to  be  valued  in  a  more  fully  developed  and



quantitative benefit-cost analysis.








5.1  Ozone-related Changes in Morbidity



     Two  points should  be  made  clear  at  the  outset  of this  exercise.



First,  because of  their number  we  cannot  calculate health  improvements




using  every one of  the  models  estimated in  Chapter 4.   Rather, we are



forced to choose from among  them when making illustrative estimates of the



changes  in  acute and  chronic morbidity that might  accompany changes  in



ambient ozone concentrations.  We have deliberately chosen from among those



models  where   ozone  concentrations   were  positively  and  significantly



associated with acute or chronic illness".



     This does not necessarily maximize the size of the estimated change in



health status accompanying a hypothetical change in ozone, since larger but



not  significant  coefficients can often  be found among  the models  we have



estimated.  But  it  should be recalled that  although  the associations used



in this  chapter  are  all  significant at  at least the  10  percent level, the



estimated  coefficients   on  ozone—while  virtually  always  positive—were



insignificant as often as not in the  analysis presented  in Chapter 4.  In



fact, this  was  the  rule rather than  the exception in our investigation of



ozone and  chronic  respiratory and  cardiovascular  disease.  To  repeat,  we



have  deliberately  chosen to illustrate the  application of our  estimates



using  models   where  positive  and  significant  coefficient estimates  were



obtained.

-------
                                    5-3


     The  second  point is  of even greater  importance.   It deals  with the

most critical, but  also  the most often  overlooked,  caveat associated with

the  interpretation  of regression results:   these results can demonstrate

correlations (or associations) but cannot prove  causation. Thus,  when one

observes  that  ozone or smoking  or  some other regressor  is positively and

significantly  associated  with  morbidity,  one  can  say  that  this  is

consistent with the hypothesis that the relationship is causal, but it does

not  "prove"  causation.   This is important  to what  follows  because using

regression results  to  estimate changes  in morbidity resulting from changes

in ambient  ozone  can  only  be done  with confidence  if  the relationship is

causal.    While  clinical and  toxicological  studies  have  provided  much

evidence to suggest why and how such relationships could be causal, readers

should  keep  in  mind that  our  results  are  only   consistent  with  this

hypothesis.



     5.1.1  Acute Morbidity

     One  of the  areas where  positive  and  significant  associations  were

identified between  ozone  and  acute morbidity was minor restricted  activity

days among adults,  due to  all  causes.   It  is fairly straightforward to use

several  of the  models finding  this association to  estimate changes  in

health status resulting from  changes in  ambient  ozone.   Acute Model 1 from

Table 4-1 can be expressed as:

                                     /\
(1)    MRADs = .772 -i-  2.48 03NR01 + 26.X,                , where
                                     1

-------
                                    5-4






               MRADs  =  number of minor restricted activity days due to



                         all causes for a given individual during the



                         two-week reference period



               03NR01 =  average daily maximum 1-hour ozone concentration



                         during that period (in ppm) at the monitor nearest



                         the individual's home

                /\


                ^iXi  =  tne otner independent variables and their



                         estimated coefficients.



Differentiating  (1) with respect to ozone, one gets:




(2)    3 (MRAD )/3(03NH£)1) = 2.48
             3


     In  other words,  a one  part-per-million  increase  (decrease)  in  the



average  daily  maximum  1-hour  ozone  concentration  during  a  two  week



reference  period  would  mean  2.48 additional   (fewer)  minor  restricted



activity days due  to  all  causes  during that period for each individual.  A



more meaningful  way  to  interpret  this  effect  results from multiplying 2.48



by  10    since ozone concentrations are usually expressed  in  hundredths of



parts-per-million.  Thus  each .01 ppm change  in  the average  daily maximum



would  result  in  .0248  additional  or fewer  MRADS  per  person  during  a



two-week period.  Multiplying  by 26 to  convert  this to  an  annual figure



suggests that each  .01  increase  (decrease) in the average  daily  maximum



1-hour ozone concentration will result in 0.64 additional (fewer)  MRADs due



to all causes per  person-year.  Extrapolating  to  a population of about 110



million  adults   (persons _>  1? years) in  SMSAs in the United  States,  this



implies 70.4 million additional (fewer) all-cause  minor restricted activity

-------
                                    5-5






days  per  year  if  the  average   daily maximum  ozone  concentration  was



increased (decreased) by 0.01 ppm in each of these SMSAs.



     This estimate  can be compared  with one derived  using another of the



models we  estimated.   Consider,  for instance,  Acute  Model (5)  from Table



4-1, where  ozone was measured  by the square of  the  average daily maximum



(the coefficient on the  ozone variable had a higher t-3tatistic  in this



model than in Model 1).  Using Model 5, the  partial derivitive is:




(3)   3(MRAD)/3(03NB01) = 2 x  16.7 x 03NR01



Evaluating ozone at the mean of the sample (.045), the change in the number



of all-cause MRADs  per  person-year for each  0.01  ppm change in the average



daily maximum reading is?



(4)    2 x 16.7  x 0.045 x .01 x 26 = 0.39.



Applying this figure to the same adult SMSA  population used above yields an



estimated change in all-cause MBADs of 42.9 million per year for a 0.01 ppm




change in the average daily maximum  1-hour ozone concentration.



     The difference between  these two illustrative  estimates is due simply



to the difference in the two estimated coefficients.  An examination of the




ozone coefficients  in Table  4-1,  where these two  models  of acute mordibity



were originally  presented,  indicates  that still  different estimates would



have  been  obtained   had  alternative  models   been  selected  for  these



calculations.



     Another  area  where  we  often   found  ozone  to  be  positively  and



significantly associated with morbidity was  acute respiratory disease among



adults.  We can  make estimates similar to those above using models in which



the dependent variable  is the  number of total restricted activity days due

-------
                                    5-6






to respiratory  disease  during the two  week reference period.   (Recall we



found   no  statistically   significant  association   between  ozone   and



respiratory-related work loss or bed disability  among  adults.)   An example




is Model  48 where  the coefficient on ozone  is  positive  and  significant at



the 5 percent level.  From that equation, it can be seen that:



(5)  3 (TRADRSP)/3(03NH£>1) = 1.54



     Thus,  to  obtain  an  estimate  of  the  number  of  additional  total



restricted activity days  per  person-year due to respiratory  disease  for a



0.01 ppm change in the average daily maximum ozone concentration, one makes



the  following simple calculation:   1.54 x  . 01  x  26  = 0.39.   A  .01  ppm



change  in the average  daily  maximum  1-hour  concentration of ozone  would



therefore imply a change of 42.9 million total restricted activity days due



to respiratory disease  alone  each year, assuming an adult SMSA  population



of 110 million in the United States.








     5.1.2  Chronic Morbidity



     The results in Chapter 4 can  also  be used  to estimate the effect of a



change in the average annual hourly  ozone concentration  (as  opposed to the



average  daily maximum)  on the likelihood  of chronic respiratory  disease



(CRD).   In  this  exercise, it is useful to work  with the elasticity of the




probability of  CRD with  respect  to  ozone  concentrations,  defined as  the



percentage  change   in  the  former for  a given  percentage  change  in  the



latter.  Consider first  the general form of  the estimating equation used in



our analysis of CRD.



It is:

-------
                                    5-7

              >PCRD  -N        n                 where
                r-)-a +  z*lxi
                 CRD        i-1
           PCRD     =  fcne Ppobabilifcy tnafc a11 individual  has  a  CRD
          , . . . . ,1    =  ozone and the other independent  variables .
Defining np Q3 as OPCRD/3(03))  (03/PCRD),  we  can  differentiate  (6) to get:
(7)   np>03 = n-PCRD)ei <°3">                    where
             /N
             8^     =  the estimated coefficient on ozone  in  the regression
                       explaining PCRD.

        ?CRD' °3    =  tne 3amPle means  of  PCRD  and 03.
     Using chronic Model  9 from Table  U-14  to  illustrate,  it can be seen
that when  (7)  is evaluated at  the  means of ozone and  the  probability of
CRD, then np Q^ = ( .92) (13.9) ( .02) = 0.25,  since ?CRD  =  .08 and ~0~3  =  .02 in
the  sample over  which Model  9 was  estimated.   Using  this  to  estimate
changes  in the  numbers of  individuals with CRD, consider a  10 percent
increase  (decrease)  in the average  annual  hourly ozone  concentrations in
the  SMSAs  of  the United States.   Since n       -  0.25,  this  implies a 2.5
                                          P , Co
percent  increase (decrease) in the  probability  of  CRD, or  an  increase
(decrease) from approximately   8  percent to  8.2  (from 8 to 7.8  percent)
percent of the  population.   Since there are roughly  110  million adults in
SMSAs in the United States, this implies an eventual increase (decrease) of
220,000  cases  of chronic  respiratory  disease  at a  given  point  in time.
Depending  upon  the model  used  to determine the coefficient  on ozone, this
estimate could  be higher or lower.   For instance,  if  chronic Model 16 from

-------
                                    5-8






Table 4-14 had  been used, the estimated change  in  numbers  of cases of CRD



for any change in ozone would be higher by about 60 percent.  On the



other hand, using coefficients from models  10  or 13 would cut the estimate



above in half.



     To summarize, if_ the coefficients used in the analysis in. this and the



previous section accurately reflect the association between ozone and acute




and chronic  respiratory and  other disease, and  if the  relationships  are



causal,  they  imply  potentially  significant changes  in acute  and chronic



morbidity for changes in  ambient  ozone concentrations.   For  instance,  a



hypothetical  change of  0.01 ppm  in  average  daily  maximum  1-hour  ozone



concentrations implies a  change  of several tens of millions in the number



of  minor  restricted activity days due to  all  causes among  adults,  and  a



similar change in the number of total restricted activity days due to acute



respiratory  disease.    A  10 percent  change  in  the  longer-term  ozone



concentration,  measured   by average  annual  hourly  readings,  implies  an



eventual 2.5  percent reduction  in  the likelihood of  CRD,  or about 220,000



cases at 1980 population levels»








5.2  Acute and Chronic Illness;   Feedback Effects



     Even  if  ambient  ozone had no  direct effect  on acute  morbidity,  it



might influence  it  indirectly via  its possible effect  on chronic illness,



which our results suggest  is strongly associated with  acute disease.  For



instance, in  acute  Model  1 from Table 4-1  (used above  to estimate changes



in  all-cause  MRADs  due  to  changed  ozone  levels)  the  coefficient on  the



dummy variable  indicating  an individual  has a chronic  illness  is positive

-------
                                    5-9





and significant at greater  than  the  1  percent level.  It is useful to take



these possible effects into account in making illustrative estimates.



     A simple model may illustrate this point.  Suppose




(8)  S = S* + SC,


           A       C
where S,  S , and  S   represent  respectively  total days spent  sick during



some period,  days  spent sick due  to  acute illness, and days ill  due  to a



chronic illness (recognizing that the distinction between acute and chronic



illness is not a simple one to make).  Suppose further that:



(9a)  SA = f(X,Z,SC)         and



(9b)  SC = g^Y,Z),         where
X   and   Y  are   personal,   socioeconomic,   or  non-ozone   environmental




characteristics  that  predispose individuals  to  acute and  chronic illness




respectively,   and  Z  is  a  measure  of  ambient  ozone which  is assumed  to




affect  both  acute and  chronic illness.   (In actuality, acute  illness  is




probably  related more to  peak than average  concentrations,  while chronic




disease  may  be  more  influenced by cumulative  dose and  therefore  more




closely related  to average annual concentrations .  We could link peak and




average concentrations through  some  complex  relationship in  this  model but




choose the present formulations for simplicity. )




     Differentiating  (8) with respect to Z, we obtain:




(10)  3 S/3Z = f z + Sz(fsc +1).



The  term  f~  in  (10)  is  the  direct  effect of ozone on acute  morbidity,  as



estimated in our models-  of MRADs, TRADs,  WLDs,  and  BDDs  in Chapter 4.  The




second  term,  g^tfgC  +  1),  represents  the  potential  direct  and  indirect




effects of ozone on total sickness operating through chronic illness.

-------
                                   5-10





     For  an illustrative  quantitative estimate  of  the magnitude  of the



combined  effect, note  that in acute Model  1  a 10 percent  increase in the



average  daily  maximum  1-hour  ozone concentration  (from  a mean  value of



.045) would add to  each  individual's  all-cause MRADs  during  a two-week



period in the following way:



(11)   3  (MRAD)/3(03NR01) = .0248 x .0045 = .011



Thus, each  adult SMSA  resident would suffer  a direct increase in MRADs of



0.29 per  year  (=26  x .011).  For an adult SMSA population of 110 million,



this would amount to an additional 31.9 million all-cause MRADs per year.



     If  average  annual hourly  ozone concentrations  also  increased  by 10



percent, then from chronic Model  9 in  Table  4-14 it can be determined that



the likelihood of chronic respiratory disease will increase by 2.5 percent,



since n      -  0.25.   As  demonstrated  above, this  implies an eventual
        r ,Uj


increase  of  220,000  cases  of chronic respiratory disease  each year.  From



acute Model  1  (Table 4-1),  the  effect  of  this change in chronic illness on



acute morbidity  as measured  by  all-cause  MRADs can be determined.   The



coefficient  on CHRLMDUM (the dummy  variable  indicating the  presence  of a



chronic  illness) is  approximately 1.30, indicating that  a chronic illness



induces  33.8 additional  all-cause MRADs each  year  (26  x 1.3).  Since each



of  220,000  individuals would  be thus  affected, MRADs  would increase by



about 7.4 million  each year among the  adult  SMSA  population in the United



States.   In other  words,  the  indirect  effect on all-cause MHADs  of an



ozone-induced  increase in  the  likelihood of chronic  respiratory disease



would amount to about 23 percent (7.4 x 106 *  31.9 x  106) of the direct

-------
                                   5-11





effect.  This points out the importance of considering the possible effects



of ozone on chronic morbidity.



     Again we  must remind  the reader that the  positive and  significant



association  identified  in  Model  9  (table  4-14)  and  used  for  these



illustrative estimates is not a. robust one.   The  coefficients  on ozone in



most other  similar models of  the  determinants of  CRD  are not significant



(although always  positive).    We  use Model  9  here only to  illustrate  the



possible magnitude  of a "feedback"  effect  if one  exists, and  we abstract



from the fact that chronic illness would take some time to develop or abate



if  ozone concentrations  were  to  change.   Our results  suggest  that  the



effect may be a significant one if further research confirms  the likelihood



of its existence.








5.3  Relative Risks



     There are a number of ways one can assess the credibility of empirical



results.   First,  t-ratios  or  other measures of  statistical  significance



shed  light  on  the likelihood  that  observed associations are  accidental.



Next, within a  particular regression,  it  is  important  that other variables



behave   plausibly.     To  the  extent  that  many  other  variables  have



counterintuitive signs,  it casts  doubt on the validity of those estimated



effects  that  do conform to prior  expectations.   Also, the  sensitivity of



observed associations  must be assessed with respect to the  measurement of



dependent and independent variables,  the  inclusion or  exclusion of certain



regressors, the functional form of the estimating equation, the presence of



"influential" observations,  and  other changes  in the  basic approach to

-------
                                   5-12






estimation.   In Chapter 4 we  relied on all  these  means to  determine  the



plausibility of our findings.



     There  is   another  approach  to  help  determine  whether  observed



associations seem plausble, one which might be referred to  as relative risk



analysis.  Under this approach, the  predicted effects  on health of changes



in ozone  (or  other variables of interest) are compared  with the predicted




effects  of  changes  in  other  important  regressors.    If  considerable



information is available about these other risk  factors, such a comparison



can sometimes suggest the plausibility of the observed  associations between



the measure of health and the variable(s) of interest.   In  this section, we



briefly  explore  several of the models  estimated in Chapter  4  and  used in



this chapter for illustrative health benefit calculations.



     Consider'first Model 1 from Table 4-1 above.  Using this model, it was



estimated  that  an  increase  (decrease) of  0.01  ppm in the  average  daily



maximum  1-hour  ozone concentration  during the  two week reference period



would  result in  0.0248  additional  (fewer)  all-cause MBADs  during  the



two-week  period  (or 0.64 more  or less MRADs  per  person-year).  How does



this compare  with  the effects of  changes  in other  independent variables?



According  to  Model 1,  a change  in smoking status—from being  a  never- or



former smoker to being a current  or occasional  smoker—would  add  0.15 to



the  number  of  all-cause  MRADs   during  a  two  week  period  (a  finding



significant at  the  5  percent level).   This  would be  equivalent to  3.9



additional  all-cause MRADs  per person-year, or  six times  the  effect  of a



0.01 ppm increase in the average daily maximum ozone reading.

-------
                                   5-13





     It  ±s  not  surprising  that   the  hypothetical  change  in  ozone  is



predicted  to have  a  smaller effect  on  acute  illness than  a  change  in



smoking status, although  one might have expected  the relative  size of the



ozone effect to be  somewhat smaller.  That  it is not may be due  to the



crudeness  of the  smoking  dummy  variable,  discussed  at   some   length  in



Chapter 4.   By contrast to  the  effect  of smoking, Model  1 from  Table 4-1




suggests that the presence of a  chronic limitation would add  1.3 all-cause



MRADs  to   each  two-week  period,  or  an  additional 33.8   person-year



(significant at the  1 percent level).  This  effect is more than 50 times



the effect of a .01  ppm increase in ozone.   If Model 5 (Table  4-1) is used



to make such comparisons,  the difference between a change in smoking status



and  a 0*01  change  in the  average  daily maximum ozone  concentration  is



larger.  In  this case, a  change  in smoking status again adds  3.9 all-cause



MRADs to  an individual's  yearly total, as  compared to  an additional 0.39



from  a  hypothetical  0.01   ppm  change  in ozone.   Thus,  the  predicted



difference is an order of magnitude.



     Turning to acute respiratory disease, Model 48 (Table  4-7)  can be used



as  a  basis for  comparing  relative risks  as  estimated   in  this  study.



According to that model, a 0.01  ppm increase in average daily maximum ozone



concentrations  would  result in  an additional   0.0154  total  restricted



activity  days  per  person due  to  respiratory disease  during  a  two-week



period, or about 0.4 per person-year.  Model 48 also suggests  that a change



in smoking status like that discussed above (in which SMOKY1ND  takes on the



value 1 rather  than  0) would add  0.03  TRADs per  person due to  respiratory



disease to each two-week period,  or 0.78 per person-year.  In  this example,

-------
                                   5-14




the change in smoking status is  only twice  as harmful as a 0.01 ppm change



in  ozone, even though  the explicit focus  in  Model  48 is  respiratory



disease—as opposed to acute morbidity due to all causes as in Models 1 and



5.  Using Model 48, a chronic  illness is predicted  to  have  sixteen times



the effect on respiratory-related TRADs as the illustrative 0..01 ppm change



in ozone.



     It  is  useful  to  perform  this  same  experiment  to  put  in  some



perspective our  findings  about  ozone and chronic respiratory  disease.   To



do so, consider Model 9 from Table 4-14 as a basis for such a comparison, a



model which  relates  each of the independent variables  to  the probability



that  an  individual will  have  a chronic  respiratory  disease.   In  this  as



well  as  in the  other models explaining chronic  illness,  ozone is measured



by  the average  annual hourly concentration  rather than by the average  of
                                        i


the highest daily 1-hour readings, which were used in the analysis of acute



morbidity.  This is important because for the 14,441 adults that constitute



our overall sample, the average annual hourly reading in 1979  was 0.02 ppm.



Thus,  when we discuss a hypothetical 0.01 ppm change  in the average annual


hourly reading,  it amounts to a 50 percent change from the mean.  Since the



average  daily maximum value  for these same  14,441  people  was  0.045 during


their reference period, the  0.01  ppm change  referred  to previously in this



section was only slightly more  than 20 percent of the mean.



     According  to Model  9,  an  increase of  0.01  ppm  in  the average annual


hourly ozone concentration would increase the probability of CRD by



0.0103  percentage  points  [3P/3C03ANAHAV) =  ^3(1-?^)   3   =  (.08)(.92)



(13.94)(.01)   =  .0103].    A  change of one cigarette  smoked  per  day

-------
                                   5-15






increase  the probability  of CRD  by  .00138  [3P/9(NCIGSDYN)  =  (0.8)(.92)




(.0187) =  .00138].   Thus,  an individual would  have  to smoke an additional




7.5 cigarettes per day  to  incur  the  same added  risk of CRD as is predicted




to arise from an increase of 0.01 ppm (or 50 percent) in the average annual




hourly  ozone concentration.    However,  it  should  be  observed  that  the




coefficient  on NCIGSDYN used to make  this illustrative comparison  is  not




significant at conventionally accepted levels (its t-ratio is only 0.51).




     Using Model  9,  it  can also be  seen that approximately  twenty years




would have to be added to the age of an average individual to have the same




effect  on the probability  of CRD  as would be  associated  with a  0.01  ppm




change  in ozone concentrations.   It can  also  be determined  from  Model 9




that an increase of  $5720  in an individual's  annual  household income would
                            •



have the  same effect on the likelihood of CRD  as a decrease of  0.01  ppm




ozone, measured by average hourly concentration.




     Model 16 in  Table 4-14  is  another in which  ozone  was  positively  and




statistically significantly  associated with the probability  of CHD.   Using




it  as  the   basis  for  a  comparison  of  relative  risks,   an  extremely




implausible  result obtains.  From that  model,  an increase of 0.01  ppm in



the average annual hourly ozone concentration is predicted  to increase Pr-n
                                                                        OnD


by  1.66  percentage  points.   An equivalent  increase due to  smoking would




require an individual  to  smoke  about 59 additional  cigarettes  per  day, an




increase of  three packs  (recall  this  is  based  on a parameter estimate that




was not significant).   In spite of its  representing a 50  percent increase




in  the   average  annual   hourly ozone   concentration,  it   is   obviously




implausible  that an  increase of 0.01 ppm  would increase the likelihood of

-------
                                   5-16






chronic  respiratory  disease by  the same  amount as  an increase of  three



packs  of cigarettes smoked  each day.   Model  16 also  suggests that  the



hypothetical  0.01  ppm  increase  considered  here would  be equivalent,  in



terms of increasing the likelihood of CRD,  to adding forty years to  the age



of  the average  individual or  subtracting $11,000  from annual  household



income (the  coefficients  on  both age and income were  significant at  the 5



percent level in Model 16).  Table 5-1  presents these relative risks.



     To  summarize,  in the  case  of acute morbidity  due to all  causes  (as



measured  by  MRADs),  our calculations  suggest  that  approximately  a  20



percent  increase in  the average daily max.mum ozone  concentration  (from a




mean  of  about  0.045  ppm) will  add  about one-sixth  to one-tenth as  many



MRADs  to an  individual's  predicted  total as a  change  in that individual's



smoking  status   (from  never or  former  smoker  to   currant  or  occasional



smoker).    The  coefficients  on  both  ozone   and   smoking   status   were



significant  and had the expected  signs  in  the  Models on which  these



estimates  were  based.    While  we might  have expected  the  relative  risk



associated with smoking to be larger, the smoking variable we  employed left



much  to  be desired.   It  may bias downward the  predicted  effect of  smoking



on health, and thus affect our  comparisons  of relative risks.



     Shifting to chronic  respiratory disease and its  correlates, we  found



using  one  model  that  an  increase  in  the  average  annual  hourly  ozone



concentration  of  0.01   ppm (a  50  percent  change   from  the  mean)  was



equivalent to smoking an additional 7 1/2 cigarettes per day in terms  of

-------
                                   5-17
Table 5-1.  Changes in Selected Independent Variables Required to Change
            the Probability of Chronic Respiratory Disease by the Same
            Amount as a Change of 0.01 ppm in the Average Annual Ozone
            Concentration

Variable (and description)                           Change Required


o  NCIGSDYN (Number of cigarettes                           59
             smoked per day)

o  AGE (age in years)*                                      40

o  INCOMCON (annual household income                      11,000
             in dollars)*
     •Coefficient significant at 95 percent level.

     1Using Model 16 (Table 4-'

-------
                                   5-18






increasing the likelihood of  CRD.   Using another model  in which ozone was



positively  and  significantly  associated  with  CRD,  we  found  that  an



individual  would  have  to  smoke  an  additional  three  packs  per day  to



increase his or her risk of CRD  by the same amount as it would be affected



by an increase of  0.01 ppm.   This  latter result is highly implausible.  It




serves  as  a reminder  of  the  problems inherent  in  selecting  somewhat




unrepresentative results to make these sorts of comparisons, not to mention



making extrapolations based on insignificant parameter estimates.








5.U  Monetary Benefit Estimates



     Although we  do not derive  monetary estimates  of nealth  benefits  In



this report, the epldemiological findings in Chapter 4 and the illustrative



calculations in  the previous  sections of  this  Chapter could  be  used for



that purpose.   In this section  we suggest how.   This section  is no more




than suggestive, however.  RFF is just beginning work on the identification



and  valuation  of  the  benefits and  costs  of  alternative  ozone standards.



More rigorous analysis of human health benefits, possibly using the present



epidemiological  study as  well  as many  other epidemiological,  clinical,



and/or  toxicologlcal  studies, will  go  into  the  valuation  Issue  in much



greater detail.



     The  MAAQS  for  ozone  is  expresed in  terms  of the  maximum allowable



concentration  (0.12  ppm)  for the  second  highest  hourly  reading  at  a



particular  monitor in a  given year  (actually the standard  is not  to  be



exceeded more than three times In  three years  on an expected value basis).



However,  in  our  analysis  of  acute   and  chronic  morbidity,  ozone  was

-------
                                   5-19





measured, respectively,  by the average  of  the daily  1-hour  maxima during



the  two-week reference  period,  and  by the  annual  hourly  concentration



averaged  over  1979 or  the  period  1974-79.    Thus,  the first  step  in



estimating monetary health benefits using  the dose-response  relationships



in this  report  would involve  linking  ozone as currently  measured for the



standard to the way it is characterized in our study.



     There are  a  number  of ways this  could  be done, only  two of which are



discussed here.   Consider  first a simple "proportional rollback" approach.



Under such an approach, if a hypothetical change in the ozone standard from



0.12 ppm to  0.14  ppm (or about 16.7 percent) were under consideration, one



would  assume that  all  ozone  readings  would  be  shifted   up by  the  same



percentage.    Thus,  if   the  average   daily  maximum   1-hour reading  was



currently 0.06  ppm  in  a  particular area, the new value would be assumed to



be  0.07  ppm.   This increment  of  0.01  ppm  could  then be  used  to predict



additional  all-cause MRADs  or  respiratory TRADs   in the case  of  acute



illness.  Similarly, if  the  average annual  hourly concentration in an area



was  0.024  ppm prior to the  hypothetical  change,  it  would  be  assumed to be



0.028 after the relaxation.  This change would then be used to estimate the



altered probability of chronic respiratory  disease  if  it was  determined to



use  our  ambiguous findings about ozone  and  CRD  as  the basis  for a benefit



calculation.    The   same  exercise  could be  repeated  for a hypothetical



tightening of the ozone  standard,  say  from  0.12 ppm to 0.10  ppm.  Again, a



uniform percentage  change could be assumed.



     A  more  complicated  and sophisticated  approach could also  be taken.



This would  involve statistical estimation of  the  relationship between the

-------
                                   5-20






second highest  hourly reading at  a  given monitor during a  given year and



the two measures  of  ozone used in our  analysis—the  average daily maximum



(in  the   case   of  acute  morbidity)   and  the  average   annual  hourly



concentration (in  the case of chronic  morbidity).   Using the  1979 hourly



data  from  all  the  ozone  monitors  meeting  certain  reporting  criteria,



several higher-order polynomials as well as other functional  forms could be



estimated  to  determine which best fits  the data.  The  resulting equation



could then be used to predict  changes in the ozone concentrations used for




the dose-response estimation given changes in the standard.



     This  would have one  advantage  compared to the  proportional rollback



approach:   it  would allow for  the  likely  possibility that increases  or



decreases  in allowable  extreme  values  (like  the  second   highest  hourly



reading in a  year) would  not  imply identical percentage changes throughout



the whole distribution of hourly values.    If  extreme  values  are  to  be



reduced, for example, it is possible that the mean of the distribution—the



average annual  hourly average  in this case—would be  changed very little.



This  more sophisticated  conversion  technique  would  be a  more  realistic



means of  anticipating the efects  of a change in the  national  ambient air



quality standard.



     Other  than to  point  out  that  predicted changes in air quality must



then be linked  to exposed  populations in order to determine health effects,



we say nothing  here about  population exposures.  Determining populations at



risk depends  critically on the assumptions made in a benefits analysis—for



example,  if  standards are tightened, do areas  currently experiencing air



quality better  than the new standards "drift up" to the new standards?  Or,

-------
                                   5-21






do more stringent  standards  only affect the ultimate  air quality of areas



currently  just  meeting  or  exceeding  the  old  standards?   Questions  like




these will  be  addresed in a  subsequent analysis.  Suffice  it  to say they




must be addressed if dose-response relationships like those above are to be




used to make monetary estimates of ozone-related health benefits.




     Once dose-response relationships are coupled with populations at risk,




the final step in estimating benefits in monetary terms is assigning dollar




values to  the various  types  of  resulting health changes.  Here again there




are a number of ways this could be accomplished.  All are linked in one way




or another to an individual's willingness to pay (or be compensated) for an




improvement  (or  worsening)  in  his  or  her  health status. This  is because




willingness to pay is  the conceptually correct measure of economic benefit




in welfare economics.^




     Direct elicitation  of  individuals' willingness  to  pay  for short-term




improvements  in   health  has  been   attempted  through  questionnaires.




Individuals  were  asked  how  much  they  would  pay for the  alleviation  of




symptoms  such  as  shortness  of  breath,  chest  pain, coughing and sneezing,




and  head  congestion  and closely related  symptoms.   The duration  of  the



reduction  in  symptoms was allowed  to vary in the study so  that estimates




were obtained for one  day's worth of alleviation, as well as one week's and




finally three  months' worth  of relief.   (Thus, the  resulting valuations




would be most useful in preparing monetary estimates of changes in acute as




opposed  to chronic  illness.)  Willingness  to  pay  was  elicited  for both




"mild"  and  "severe"  cases  in  each symptomatic   category.     Average



willingness to pay of  those  surveyed varied  between $2.31 to avoid one day

-------
                                   5-22






of minor  coughing  and sneezing to  $97.80 to avoid three months  of severe




shortness of breath (at what appear to be 1975 prices and incomes).




     Like  all  estimates obtained  in  surveys,  these are  subject  to  hard




questions about sample  selection  bias, as well as  bias  resulting from the




way the survey  was  designed and administered.  We  mention  these estimates




here because the symptoms for which valuations were elicited are similar to




some of  those thought  to  result from exposure to peak  concentrations of




ozone.  Therefore, they might be useful in developing a range of values for




the health changes obtained using the dose-response relationships estimated




in the "acute" portion of our study.




     Concerning our estimates of the effect of ozone on acute morbidity, an




interesting valuation problem arises.  Recall that we found no relationship
                        •



between ozone and  either bed  disability days or work or school  loss  days




among the adults and  children  analyzed.   But we did find frequent positive




associations between  ozone  and minor  restricted  activity days due  to all




causes  and  total  restricted  activity  due to  respiratory  disease  (in




adults).  Yet these restricted activity days may be more difficult to value




than  the  more  serious   types of  acute morbidity because  they  are  more




difficult to conceptualize and define.




     For  instance,  an  individual  might be  willing to pay  very  little to




avoid an  acute  illness  that prevents  him from  doing  a minor but necessary




and uninteresting chore, yet would be willing to pay much more if this same



illness  or  condition  prevented  him  from  taking  part  in  an  eagerly




anticipated  social  or  recreational  event.    Yet  both would  be  considered




restrictions  in regular activity  under the  HIS  protocol.    This suggests

-------
                                   5-23






that even survey data on willingness to pay would be difficult to interpret




and  use  unless  questions   were   tied  to  specific  types  of  restricted




activity.  Moreover, since restrictions on  activity as  defined by HIS will




generally not  entail  out-of-pocket  spending or  lost earnings,  even this




cost-of-illness proxy for willingness to pay will be of little assistance.




These factors  suggest  that  a range of values should be used in converting




restricted activity days to monetary benefits.




     Difficulties arise in valuing chronic  illness, as  well, although they




are  not  of  as  much  consequence  for this  report  because  of the  very




tentative nature  of our  findings  regarding ozone  and  chronic respiratory




disease.  There are some survey data on willingness to pay to avoid chronic




illness comparable to  those  cited  above for acute morbidity.^  These data,



which suggest that individuals would be willing to spend as much as several




hundred  dollars  per year to  be  free of  chronic  illness  related  to  air




pollution, must  be  treated very carefully—individuals  can  be expected to




have much more difficulty assigning meaningful  values   to long  periods of




time free from chronic respiratory or  other kinds of disease than to a day




or  week  free  from coughing and  sneezing.    Among  other  things,  such



estimates would depend heavily on the prior health histories of individuals



surveyed.   In  addition,  estimates  will  be sensitive  to the  sample upon




which they  are based,  as well  as  upon a number  of the characteristics of




the survey itself and  the way it is administered.




     Estimates have  been  made of the annual costs  associated with various




types of  chronic  illness.  One study reported  the annual costs associated




with  CUD comprised  of  (i)  a  "direct"  component  consisting of  doctors'

-------
                                   5-24






charges plus the coat of hospitalization  and/or  nursing home care, (ii) an



"indirect" component reflecting lost earnings,  and (iii) a component called



"indirect mortality cost" which is  the  product of the probability of death



from the  disease times  the  earnings that  would be  lost  in the  event of




death.    By disease,  these  annual  costs were  estimated  to be:   chronic



bronchitis,   $237;   bronchiectasis,   $701;   emphysema,  $1294;   chronic



intestitial pneumonia,  $195 (all in $ 1981).



     For a number of reasons  these estimates would not be particularly good



proxies for individuals' willingness to pay to reduce the prevalence of the



diseases  considered  in  our  study.    First,  like  all  cost-of-illness




estimates, they ignore  both the pain and  suffering associated with illness




as well  as any  defensive  measures  individuals  may have taken  to protect



themselves against chronic illness.   Second,  to the extent  that "indirect



mortality cost"  is  an  appropriate component of  such an estimate—-and this



is unclear—it should be based not on foregone earnings during the extended



years  of  life,  but rather  on willingness to  pay  for reduced mortality



risks.  Nevertheless, if used with care these and other recent estimates of



the costs associated with  chronic respiratory  disease might  be of some use



in making monetary estimates of the health benefits associated with



alternative NAAQSs for ozone.

-------
                                   5-25



                          Footnotes to Chapter 5



      See Just, Hueth, and Schmitz [1982]  or Freeman [1979].


     "Tor example, see Loehman et al.  [1979].


      See  Brookshire, d'Arge,  Schulze,  and  Thayer [1979]  and  Loehman,


Boldt, and Chaikin [1981].

     4
      See d'Arge, Kask, Case, Ben-David,  Eubanks, and Anderson  [1982],  pp.


39-43.


      See Appendix  C  to  this  report  for an  analysis  of the effects such


omissions can have on benefit estimates.

-------
                                   5-26


                          References to Chapter 5


Brookshire, David, Ralph d'Arge,  William Schulze, and Mark Thayer,  "Methods
     Development  for Assessing  Pollution  Control  Benefits,"  Volume  II,
     "Experiments in Valuing Non-market Goods:   A Case Study of  Alternative
     Benefit Measures of Air Pollution Control  in the South Coast Air  Basin
     of Southern  California,"  Publication no.  EPA-600/5-79»001b,  February
     1979.

d'Arge, Ralph, Susak Kask, James Case,  Shaul Ben-David, Larry Eubanks,  and
     Curtis Anderson, "Air Pollution and Disease: An Evaluation of the  HAS
     Twins,"  a  report  to  the Environmental Protection  Agency, Office  of
     Research  and  Development  under  contract  no.  CR-808893-01   to  the
     Resource  and   Environmental   Economics  Laboratory,  University   of
     Wyoming, 1982.

Freeman, A.  Myrick,  The Benefits of  Environmental  Improvement  (Baltimore,
     Md.:    Johns  Hopkins University  Press  for Resources  for the  Future),
     1979.

Just, Richard, Darrell Hueth and Andrew Schmitz,  Applied  Welfare Economics
     and Public Policy (Snglewood Cliffs, N.J.:  Pr entice-Hal. 1), 1"9~52T

Loehman, E.,  S.V. Berg, A.A.  Arroyo, R.A.  Hedinger-,  J.M. Schwartz, M.E.
     Shaw, R.W. Fahien,  V.H.  De, R.P.  Fishe,  D.E.  Rio, W.F. Rossley,  and
     A.E.  Green,  "Distributional Analysis of Regional  Beneits and  Costs of
     Air   Quality  Control,"   Journal   o£  Environmental  Economics   and
     Management,  vol. 6 (1979), PP-  222-243.

Loehman, Edna, David Boldt, and Kathleen Chaikin,  "Measuring  Benefits  of
     Air  Quality  Improvements  in  the  San Francisco Bay  Area," report
     prepared  by  SRI  International  for  the  Department  of  Economics,
     University of Wyoming. May 198K

-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1fpT4T50N/V85-3583
4. TITLE AND SUBTITLE
Ambient Ozone and Human Health: An Epidemiological
Analysis Volume I
7. AUTHOR(S)
Paul R. Portney and John Mullahy
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Resources for the Future
1616 P Street N.W.
Washington, DC 20036
12. SPONSORJNG AGENCY NAME AND ADDRESS
U.S. Enviornmental Protection Agency
Office of Air Quality Planning and Standards (MD-12
Research Triangle Park, NC 27711
15. SUPPLEMENTARY NOTES
Project Officer: Thomas G. Walton
3. RECIPIENT'S ACCESSION NO.
5. REPORT DATE
Sept. 1985 (Date of Preparation
6. PERFORMING ORGANIZATION CODE
8. PERFORMING ORGANIZATION REPORT NO.
10. PROGRAM ELEMENT NO.
12A2A
11. CONTRACT/GRANT NO.
68-02-3583
13. TYPE OF REPORT AND PERIOD COVERED
Final report
>) 14. SPONSORING AGENCY CODE
OAQPS

16. ABSTRACT
This report is the first volume of an analysis of the relationship between
ozone and human health benefits.
•
17. KEY WORDS AND DOCUMENT ANALYSIS
a. DESCRIPTORS b.lDENTIFI
Benefit Analysis
Air Pollution, 0,
Epidemiology
18. DISTRIBUTION STATEMENT 19. SECURI
Uncle
Release Unlimited 20. SECURI
Uncle
ERS/OPEN ENDED TERMS c. COS AT I Field/Group

TY CLASS (This Report) 21. NO. OF PAGES
issified 324
TY CLASS (This page) 22. PRICE
issified
EPA Form 2220-1 (Rev. 4-77)    PREVIOUS EDITION is OBSOLETE

-------
                                                       INSTRUCTIONS

   1.   REPORT NUMBER
       Insert the EPA report number as it appears on the cover of the publication.

   2.   LEAVE BLANK

   3.   RECIPIENTS ACCESSION NUMBER
       Reserved for use by each report recipient.

       TITLE AND SUBTITLE
       Htle should indicate clearly and briefly the subject coverage of the report, and be displayed prominently. Set subtitle, if used, in smaller
         oe or otherwise subordinate it to main title. When a report is prepared in more than one volume, repeat the primary title, add volume
         mber and include subtitle for the specific title.

   5.   REPORT DATE
       Each report shall carry a date indicating at least month and year.  Indicate the basis on which it was selected (e.g., date of issue, date of
       •l>i  oval, date of preparation, etc.).

   6.   PERFORMING ORGANIZATION CODE
       Leave blank.

   7.   AUTHOR(S)
       Give name(s) in conventional order (John R. Doe, J. Robert Doe, etc.). List author's affiliation if it differs from the performing organi
       zation.

   8.   PERFORMING ORGANIZATION REPORT NUMBER
       Insert if performing organization wishes to assign this number.

   9.   PERFORMING ORGANIZATION NAME AND ADDRESS
       Give name, street, city, state, and ZIP code. List no more than two levels of an organizational hirearchy.

   10.  PROGRAM ELEMENT NUMBER
       Use the program element number under which the report was prepared. Subordinate numbers may be included in parentheses.

   11.  CONTRACT/GRANT NUMBER
       Insert contract or grant number under which report was prepared.

   12.  SPONSORING AGENCY NAME AND ADDRESS
       Include ZIP code.

   13.  TYPE OF REPORT AND PERIOD COVERED
       Indicate interim final, etc., and if applicable, dates covered.

   14.  SPONSORING AGENCY CODE
       Insert appropriate code.

   15.  SUPPLEMENTARY NOTES
       Enter information not included elsewhere but useful,  such  as:  Prepared in cooperation with, Translation of, Presented'at conference of.
       To be published in, Supersedes, Supplements, etc.

   16.  ABSTRACT
       Include a brief (200 words or less) factual summary of the  most significant information contained in the report. If the report contains a
       significant bibliography or literature survey, mention it here.

   17.  KEY WORDS AND DOCUMENT ANALYSIS
       (a) DESCRIPTORS - Select from the Thesaurus of Engineering and Scientific Terms the proper authorized terms that identify the major
       concept of the research and are sufficiently specific and precise to be used as index entries for cataloging.

       (b) IDENTIFIERS AND OPEN-ENDED TERMS - Use identifiers for project names, code names, equipment designators, etc. Use open-
       ended terms written in descriptor form for those subjects for which no descriptor exists.

       (c) COSATI FIELD GROUP - Field and group assignments are to be taken from the 1965 COSATI Subject Category List.  Since the ma-
       jority of documents are multidisciplinary in nature, the Primary Field/Group assignment(s) will be specific discipline, area of human
       endeavor, or type of physical object.  The application(s) will be cross-referenced with secondary Field/Group assignments that will follow
       the primary  posting(s).

   18.  DISTRIBUTION STATEMENT
       Denote releasability to the public or limitation for reasons  other than security for example "Release Unlimited." Cite any availability to
       the public, with address and price.

   19. &20.  SECURITY CLASSIFICATION
       DO NOT submit classified reports to  the National Technical Information service.

   21.  NUMBER OF PAGES
       Insert the total number of pages, including this one and unnumbered pages, but exclude distribution list, if any.

   22.  PRICE
       Insert the price set by the National Technical Information Service or the Government  Printing Office, if known.
EPA Form 2220-1 (Rev. 4-77) (Reverse)

-------