U-.ited States      Office o* An Qaalitv        EPA-450/5-85-005d
Environmental Protection  Planning an: Gianaards      August 1985
Agency        Rese&rcn Triangle Par\ NC 27711
A:r
Executive Summary

Ambient Ozone And
Human Health:
An Epidemioiogical
Analysis

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

                        AN EPIDEMTOLOGICAL ANALYSIS
                      aul R. Purtnev and John Mullahv
                         Resources for ths Future
                            1616 P Street, N.W.
                          Washington, D.C.  20036
                                 June
Submitted to the Economic Analysis Branch, Office of Air Quality Planning
and Standards, Envirorir/iental 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.

-------
                             EXECUTIVE SUMMARY






     In June of 1982 Resources for the Future (RFF) began a project  for  the




Economic Analysis Branch of the Environmental Protection Agency's Office of




Mr Quality  Planning  and Standards  (OAQPS).   The  purpose  of that  project




was to provide information to OAQPS on the human health benefits that might




be  associated  with  possible  alternative  National   Ambient  Air   Quality




Standards for  ozone,  one of  six  so-called "criteria"  air  pollutants.   In




September of 1983, RFF completed work on a two-volume Draft Final Report on




that  project,  "Ambient  Ozone  and  Human  Health:    An  Epidemiological




Analysis."     Since  that  time RFF  has  conducted additional  analysis in




response  to  comments, criticisms, and  suggestions arising  from the draft




report.   Many  of  these  comments were  offered  at  a  Public  Peer Review




Meeting held  in Raleigh,  N.C.  on April  3,  1984.  This  Executive  Summary




briefly reviews  the methodology and  conclusions  in Volumes  I  and  II,   the




original  draft  report,  as well as  those  in the  recently completed Volume




III, the sequel to RFF's initial work.








Volume I




     Volume I of the  1983 Draft Final Report presents the data, methodology




and  findings  of  RFF's   original  epidemiological analysis.    Chapter 2




discusses in  great detail  the  health and socioeconoraic data  used in   the




study,  the  air  pollution measurements  used  to   link  exposure to health




status, and the meteorological and other data that were also used.   Chapter




3 discusses  the  methodology used to  explore possible  links  between ozone

-------
and other pollutants  on  the one hand  and  acute and chromic Illness on  the




other.   Neither  chapter is reviewed  in detail  in  this Executive Summary.




Rather,  we  prefer to  review  here the findings of  the "Results" chapter,




Chapter U.  However,  a brief review of chapters 2 and 3 may prove useful.




     The health and socioeconomic  data upon  which  the analysis in Volume  I




is  based  come from  the 1979  Health   Interview  Survey  (HIS)  of  more than




110,000 adults and children drawn randomly  from the United States  (Volume




I,  pp.   2-1  through  2-21).   For  each  individual  included  in  the HIS,




information is elicited on personal characteristics  (age, race, sex,  etc.),




occupational or  educational status,  and  on  any of  several types of acute




health episodes  during the  two weeks  immediately  prior to the week of  the




interview.   Also included  is  information on  the presence  or absence, as




well  as  the  severity,  of any  chronic  illnesses from which the respondents




may suffer.   For  all acute and  chronic  illnesses,  coding in  the HIS is




according to the  standard  International Classification of Diseases.  Thus,




one can identify from  the HIS,  for instance, the number of  "bed disability"




days  a  respondent  suffered on  account  of  respiratory  disease  during  a




two-week  period  in  1979.    Similarly,  one  can  identify  all individuals




suffering  from  asthma,  emphysema  or  other chronic  respiratory disease.




Finally, the 1979 HIS  contained two very useful supplements.   The first  was




a  detailed  lifetime   smoking  history for  26,000  of  the  78,000   adults




interviewed  in  the  HIS.   The  second  was  a  residential  history, again




administered to one-third of the adult respondents.




     The air pollution data used  in  the RFF Draft Final Report, and in  the




more  recent  Volume III analyses,  all  come from EPA's SAROAD system  (Volume

-------
                                     3






I, pp. 2-22 through  2-46).   Hourly data were used to meaaure ozone,  carbon




monoxide,   and  nitrogen   dioxide   concentrations.      Total   suspended




particulates  and  sulfates  were  measured  on  a  24-hour  basis with  the




readings  being taken  every  six  days  generally.    Thus,  exposure  to  the




latter two pollutants may be  measured  less accurately than for  the  former.




Before matching  these  data  to individuals  (through  a process described  on




p.  2-46),  all  data were  subjected  to a  variety  of  tests  to eliminate




potentially  incorrect  values or  outliers resulting from  coding   errors.




Also described in Chapter 2 are  the procedures used to average  some  of  the




air  pollution data  over  the two-week recall  period  used  in  the  acute




epidemiological  analysis  as   well  as  over   a  six-year  period  used   to




facilitate  RFF's   chronic   epidemiological   analysis.     The   data   on




temperature,   humidity,   precipitation,   windspeed   and   other    weather




conditions come  from NOAA's  meteorological monitoring  stations around  the




U.S.




     Finally,   Chapter   2   discusses   the  sources  of   data   on   pollen




concentrations,  the  use  of  gas  stoves  in kitchen  cooking,  the   average




amount of annual paid sick leave permitted workers, and the availability  of




medical care by geographic  area in the United States.




     The measures of health  status used in the  Volume  I  analyses are,  for




acute illness, the number of restricted activity days, work loss (or  school




loss)  days,   and  bed  disability  days during  each  individual's  two-week




recall period.  The category  "restricted  activity  day"  is an inclusive one




that encompasses both work  (or school) loss and bed disability days as well




as days on which respondents had to curtail their normal activities  to some

-------
extent short of missing work  (or  school)  or bed confine-ent.  These latter




impairments we referred to in Volume I as "minor restricted activity days."



Each type of acute illness was analyzed separately.  Also separate analyses




were conducted  using acute episodes  due  to all causes,  and then episodes




limited  to  respiratory illness alone.   In  addition,  logistic regressions



were performed in which the dependent variable indicated whether or not the




respondent  had  at  least  one  day of  illness  during  the  two-week period.




Logistic regression was used exclusively in the analysis of respiratory and




other  types  of chronic  illness.    In  all cases,  children  and adolescents




(those < seventeen years of age) were analyzed separately from adults.




     Chapter 4 of Volume  I,  wherein are reported RFF's empirical findings,



represents  the  heart of  the  Draft Final Report,   The major  part  of that




chapter  is  devoted   to   the  analysis  of  acute  illness,  particularly




respiratory disease, and  its  possible  link  to short-term exposure to ozone




and  other  air  pollutants.    (Throughout  the  analysis  of  acute  illness,




exposure to  ozone  is  characterized by  the average over  the fourteen-day




recall period of  the  daily maximum one-hour reading at the monitor closest



to  the respondent's  residence;  no individuals are  included in the estimate




sample if  they  live more than  twenty  miles from the nearest air pollution




monitor.)   More  attention  is devoted  to  dose-response  estimation among




adults than among children.




     This  emphasis  on  acute illness in adults  as  related to ozone follows




from  a  strategy  adopted  throughout  Volume I:    use  sensitivity analysis



primarily to test the robustness  of any positive and significant

-------
associations discovered between ozone and illness.  Sine;  such associations




arose almost exclusively among adults and when examining acute illness,  the




report reflects this  emphasis.   This approach  reflects the fact that both




time and resources were finite in the original analysis, as they are  in  all




such undertakings.



     Generally  speaking,   and  this  is  an  important  qualification,   the




analysis of  acute illness among adults  supports  the  following conclusion:




there appears to be no consistent association between average daily maximum




ozone concentrations and  either work  loss days or bed disability days,  the




two more severe types of  acute impairments  analyzed  in Volume I; however,




it  was  frequently  the  case  that  ozone  was positively  and  significantly




associated with "minor" restrictions  in  activity,  i.e., those that did  not




necessitate  work  loss  or bed  confinement.    While there  were  few  or  no




exceptions to  the  former  conclusion,  there  were  a number of exceptions to




the  latter.   Nevertheless,  when  the analysis  concentrated on respiratory




disease in  particular—where air-pollution  health effects are most  likely




to  occur—ozone  was  regularly  found  to  be related  in  a   positive   and




significant  way  to either  the  number  of restricted activity days  or  the




likelihood of  having  at  least  one  during the  two-week recall period (see



Volume I,  Tables  4-7, 4-8 and 4-10).   For no other air pollutants was such




a_ regular  association found.




     Among the sensitivity  analyses  conducted to  test this conclusion were




the following:  (i) varying combinations of  air pollutants  other than ozone




were  included  in  the  estimating equation;   (ii)  the  square of  ozone  was




introduced to  examine  possible  non-linearities;  (iii) logistic regressions

-------
were run in place of ordinary least squares;  (iv) tighte- requirements  were




imposed  on  the  availability  of  ozone data  (a  greater  percentage  of  the




maximum  number  of  hourly  readings were  required);  (v) the  analysis  was




restricted  to  respondents  interviewed during  summer months;  (vi)  pollen




concentrations were introduced as an additional explanatory variable;  (vii)




the analysis was restricted to those suffering from chronic illness; (viii)




the twenty-mile distance cut-off  was  tightened  to ten and then five miles;




(ix) lagged values for ozone (from a previous two-week period) were used to




proxy exposure;  (x) possible  synergistic effects between  ozone  and other




air pollutants were examined;  and (xi) separate regressions were run  for a




set of "self-respondents," i.e., those who answered the  HIS for themselves.




As  suggested  above,   in  some  of  these  sensitivity  analyses   the  ozone




variable  was  not  positively  and  significantly  associated  with  acute




respiratory disease  (or  disease due to all  causes).   More often than  not,




however, it was.




     The findings  in  the  Draft Final  Report  regarding acute illness among




children  can  be  summarized   more  readily.    Virtually  no   positive   and



significant associations were  found between  ozone and  any of the types of




acute impairments  that were considered.  This  was  the case when the focus




was  on  all  types  of  illness  as  well  as  when   the analysis  focused




exclusively on acute respiratory illness.  Ozone was  associated with school




loss at  the  10  percent confidence level in one or two models, it should be




added, but never at the 5 percent level used  as the "cutoff"  in Volume  I of




the  report.   The  only  association   significant  at  the 5  percent  level




occurred  in  one model  exploring  ozone  and bed disability days  due  to

-------
                                     7






respiratory disease.  Of the other air pollutants include!  in  the analysis,




nitrogen dioxide  was  found to  be  positively and  significantly related  to




school  loss  days  among  children  in the  analysis of  illness due  to  all




causes.   While  its  coefficient  was  positive  in the  analysis  of  minor




restrictions  in   activity  and  bed  disability,  nitrogen  dioxide  was  not




significantly associated with these forms of acute illness.  The one caveat




that  should  be issued  here  is  this:   very few  sensitivity analyses were




conducted using the children's  data set because  of  the absence of  initial




findings  concerning  ozone  pollution/acute  illness   associations.    More




careful analysis  of possible air  pollution health effects among children




should be conducted.




     Volume I of the Draft Final Report also devotes considerable attention




to the possible role of  ozone  and  other air pollutants in  chronic illness.




In this analysis,  which  begins on p.  4-50,  the  focus shifts away from  the




two-week  recall   period  and  the  number   of  days  of  restricted  activity




experienced by each respondent  during  that period.   Instead, the dependent




variable  in  the  chronic  analysis is  whether  or  not the  respondent  has




chronic respiratory (or,  in some  cases, cardiovascular or other) disease.




Accordingly,  the  measurement  of ozone  and the other  air  pollutants  is  no




longer specific to the two-week recall period, as it was in  the analysis  of




acute  illness.   Rather,  ozone  and other  air pollution concentrations  are




averaged over longer periods of time.




     Generally,  the  longer  averaging time  is   one   year.    However,  a




multi-year air pollution  data  set  was created for  this project.   Thus,  in




some  of the  models  fit  in  Chapter  4,   individuals  are  matched  to  air

-------
pollution  concentrations  over  the  six  year  period  V.-  —79.    The  logic




behind  this   is  straightforward.    Chronic  illness,  if  related  to air




pollution  at  all,  is  likely  to result  from prolonged  exposure over many




years.  Frequently, a  single  year's  data are used to proxy exposure over  a




much  longer  period of  time.   However,  where a  longer-term  air  pollution




data set could be  assembled,  as it  was  in  this  study,  it is preferable to




the use of one year's  data.  As on  the  analysis of acute illness, the air




pollution data matched to an individual came from the monitor nearest  his




home,   so   long as  minimum  requirements  regarding  data  completeness were




satisfied.




     In general,  the analysis of chronic respiratory and other  diseases in




adults  revealed  no   pattern  of   positive   and   statistically   significant




associations  between illness and ozone or most of the other air  pollutants.




There  were occasional  exceptions,  some  of  which are  discussed  below, but




the results  contrast  with  the findings  about  ozone and minor restrictions




in activity due to  respiratory disease.




     In the  initial  analysis  (see  Table  4-13  in Volume I), all pollutants




were  measured using  a single  year's   data,   1979>   to   proxy  long-term




exposure.  No significant  associations  were  observed  for any  of the air




pollutants.   When  the  multi-year  data set  was  brought  into  play, the




results changed somewhat (see Table  U-14).   In some models, the  coefficient




on  ozone  was  positive  and  significant at the  10 percent  level    (see




equation (9)» for example).  In addition sulfur  dioxide concentrations were




positively   and  significantly  related   to  the  probability   of   chronic




respiratory  disease  (CRD).   This  latter  finding was consistent throughout

-------
much of  the analysis  using  the  multi-year  data set.    Among  the  adults




analyzed  in the  report,  this was  the  one  strong  Association  to  occur




frequently.  The  lack  of any  significant  findings with  respect to  ozone




persisted across a number of sensitivity tests.




     There was  one  exception to the negative  findings  regarding ozone  and




CRD.   This occurred when the  adult  sample was  divided  into three  groups




consisting of  those who had  never  smoked,  those  who  were ex-smokers,  and




those who  were  current  smokers.  Among  the never-smokers, ozone (measured




by the multi-year average) was positively and  significantly associated with




CRD  (Table 4-15,  equations  (23)  and  (26)).    Among  the  never  and  former




smokers,  sulfur  dioxide was  positively  and  significantly associated with




CRD.




     In the analysis of  CHD  among  children, there was but one positive  and




significant  association—that  concerned  nitrogen  dioxide  in   one  model




(equation  (32),   Table   M-16).   As  in  the  case of  acute  illness  among




children,  very little  sensitivity  analysis  was  conducted because  of  the




largely negative findings.




     Brief  attention was  devoted  in  Volume  I  to  chronic  cardiovascular




disease (CCD) (see Table U-17).  Explorations  here found multi-year  average




ozone  concentrations positively  but  never significantly  associated with




CCD.   Nor  were  any such  associations  identified  for  other  of  tie  air




pollutants considered.   Confidence  in these  findings was  bolstered by  the




fact that  other risk factors  identified  in  previous  studies  of CCD  all  had




the  expected  sign  and   were  highly  significant  in  the  analysis.    These




included age, race,  smoking  habits,  weight, and education.   In addition to

-------
                                    10






CCD,  a variety of chronic diseases, neoplastic and other rise, were examined




for possible links to ozone and other air pollutants.  No such associations




were found.   Once again, a caveat  to this analysis  is  that only the most




superficial tests  were  conducted.   It  is always  possible  that additional




analyses might produce different results.








Volume II




     Volume II  of the  Draft  Final  Report  requires  virtually  no summary.




Its primary purpose is  to  present the Health Interview Survey for 1979 and




the two  supplements,  one on smoking  and  the  other on residential mobility




(which is actually pp.  110-111)  of  the  HIS itself).  In addition, Appendix




B  of  Volume  II  contains  the  data  documentation  and  cross-referencing.




Appendix C  (actually the  third  part of Volume II)  does  present research




output from  the RFF  ozone project.   However, this  output  is theoretical




rather than empirical.




     Specifically, Appendix C  is  a  paper deriving the conceptually  correct




measure of the benefits of a pollution control program in a  simple model  of




consumer behavior.    In  the model,  pollution not only  affects  welfare  by




leading to illness but also enters individuals' utility  functions directly.




In  addition,   in  the model  individuals  can  protect  themselves   against



illness  by making  "defensive"  expenditures.    (these  might  be purchases  of




air conditioners  or  filters,  water  purifiers,  preventive health measures,




and so on.)   It  was  the purpose  of  this work to  see  how  the presence  of




such  defensive  or averting  expenditures might   affect the  valuation  of




health   benefits,   and   how   the   correct   valuation  compared   to  the

-------
                                    11
out-of-pocket  and  opportunity  costs  associated  with  pollution-induced




illness.



     The results  can  be summarized  as  follows.  First,  the  measure of  an




individual's willingness-to-pay  for  a pollution  reduction  contains only




derivatives of  the  dose-response function,  specifically  those relating  to




the effect  of  pollution on  health  and  defensive expenditures  on health.




Next,  the conceptually correct measure will  in general not only exceed the




change  in out-of-pocket  and  opportunity  costs  of  illness but also the sum




of  these  costs  plus  the  change  in averting  or defensive  expenditures.




Finally,  this  conclusion  holds  for several  different  variations  of the




basic model  in  which  workers are  given  paid  sick  leave, sickness affects




the wage rate,  and other changes as well.








Volume III




     The  third  volume of  work  completed   by  RFF for  OAQPS  is more  recent




than  the  first  two.   It represents  work  undertaken  since  the  April  1984




Public Peer Review Meeting and consists largely of additional analysis done




in response to  comments received at that meeting or elsewhere.




     Chapter 2  is methodological.  It discusses in some detail the kinds  of




problems that can arise when  ordinary least  squares  techniques are used  to




estimate  the likely  effects  of air pollution  on  health  outcomes.    On




account of  these problems—the  most  important of  which is  the  frequent




truncation  or   censoring  of   the   range   of  health   outcomes—alternative




estimating  techniques  are  reviewed.   These  include Tobit  and Cragg-class




models, truncated  normal estimation,  the  sample  selection  model, Poisson

-------
                                    12






and geometrically distributed measures  of health outcomes, and multinomial




techniques.




     The third  chapter of  Volume III  is quite  brief and is  designed  to




answer a specific and important question:  How would the analysis  in  Volume




I have  changed  if the  air  pollution readings  assigned to individuals  had




come not just from the  single  nearest monitor but from all monitors  within




some specified distance of the individuals' homes?




     The answer suggested by the analysis in Chapter 3  is  "Not very much at




all."   This  conclusion follows  from the simple  correlation coefficients




between the  nearest  monitor  readings and  the  readings averaged  over  all




monitors within  ten  and then twehty  miles  of  the respondents' homes.   For




ozone the correlation coefficients are 0.97 and 0.93 for the  ten and  twenty




mile averages,  respectively.   Thus,   substituting  either of the latter  two




for the nearest  monitor readings  would  make virtually no  difference  in  the




results.  The smallest  of all  the correlation coefficients thus calculated




for all pollutants was  0.82.




     Chapter  4  presents the results  of  some  substantial  reanalysis  of  the




Volume  I  findings  concerning  ozone  and  acute  respiratory  disease.    It




differs from  (and improves  upon) the  original work  in  several  important




respects.    First, Poisson  regression is  used  in place  of ordinary  least




squares.  For reasons  discussed  in Chapter 2 of Volume III,  this  is  a more




appropriate way  to model acute  health outcomes.  Second,  no  individual  was




included in the  analysis if the nearest  air pollution  monitor was  more than




ten  miles  from  his  home,  as opposed to twenty  in  the original  analysis.




Third,  the  analysis of possible non-linearities  is  more  sophisticated  in

-------
                                    13
the later  analysis.   Finally,  air pollution was  measured  not only by  the




reading at the  nearest  monitor during the  two-week recall period but also




by the  average  reading during  that period of  all monitors within ten  and




also twenty miles.   Annual average readings  were  also employed in several




of the models estimated.




     The  analysis  in  Chapter  4  supports  the   findings  in  Volume   I.




Specifically,  ozone  (however  measured)   is  found  to  be  positively  and




significantly associated with the number of restricted activity days due to




respiratory disease during the  two-week  recall  period.  No effect is  found




for sulfates, the  other air pollutant included in the Chapter 4 analysis.




Of the other independent variables, race, income,  the  presence of a chronic




illness,  and   temperature  were   also   found   to  be   associated   in   a




statistically significant way with  acute respiratory disease.




     In other findings in Chapter 4, the possibility of interactive effects




between ozone  and  sulfates  and  ozone and  temperature  are rejected.    In




addition,  no  evidence  is  uncovered supporting the idea of  a "threshold"




below which ozone  concentrations  are  harmless.   However,  the analysis does




support the  finding that the dose-response  relationship  between ozone  and




acute respiratory disease is non-linear.   That  finding (see equation 4.5 in



Table 4-4) suggests that the square root of the average daily maximum ozone




concentration  at  the  nearest  monitor   is   a   potentially important   and




significant determinant of acute respiratory disease.




     Chapter  5  summarizes  recent  research  on  the construction  of  proxy




measures for lifetime smoking  profiles.   The smoking controls used in both




the statistical analyses  in  Volume I and  in  Chapter  4 of this volume were

-------
                                    14






typically either the  current  rate of daily  cigarette  ccisumption or dummy




variables indicating whether the individual was a current, former, or never




smoker.    As  noted  at  various points  in  the analysis,  such  variables are




acceptable controls  for assessing  the possible  linkages  between current




cigarette  consumption  and  respiratory  illness.    However,   because  the




relationship  between cigarette smoking and respiratory illness  is  also




thought   to  have  lagged  or  cumulative  dimensions,   separate  explanatory




variables are required to effect controls for such relationships.




     Given this requirement,  and since no  single value  given  in the data




provides such information,  we explored the  detailed data available in the




HIS Smoking  Supplement in  order to assess  whether plausible  proxies for




lifetime cigarette consumption could be created.  Using information on peak




and current  rates  of consumption,  age,  age started smoking,  and duration




since  quits,  it  was  ascertained   that  various   plausible  controls  for




lifetime smoking could  indeed  be obtained from the HIS data.   It  turns out




that the  measures  are  created  in a manner  analogous  to how capital stock




values  are obtained from  investment  and depreciation  profiles in applied




microeconomic  research.   Chapter  5  presents  in  considerable  detail the




procedures used for creating these lifetime  cigarette consumption  profiles.




     Drawing on  the results  of Chapter  5,  Chapter 6  proceeds in several




directions to  a more  sophisticated analysis of  the  relationships between




ambient air  pollution  and  respiratory illness.   Not  only are  the lifetime




smoking profiles discussed  in  Chapter  5 used to control  for longer-term  or




cumulative relationships between cigarette smoking and respiratory illness,




but a new, and we believe  particularly interesting, measure of respiratory

-------
                                    15






illness is used.  That is, we consider in Chapter 6 the  Distinction  between




minor limitations in activity due to respiratory illness, i.e.,  those  where




the individual  is  not  confined to bed,  and  major  restrictions  in activity




due to respiratory illness,  i.e.,  those  where some bed confinement  occurs.



Moreover, we also consider in Chapter 6 the distinction used by  the




National  Center   for  Health   Statistics  between  inherently  acute  and




inherently chronic respiratory illness.




     To  assess  econometrically the  relationships  between  the  explanatory




covariates  and  the  new  measures   of   respiratory  illness,  we  use  a




multinomial logit  estimator  of  the determinants of the various  respiratory




illness outcome probabilities.  Like  Chapter  4, it is found that nonlinear




transformations of the ozone variable, as well as  of the smoking controls,




give superior  results.    Ozone is found to  be positively related  to the




probability of  all the  respiratory illness outcomes,  and the relationship




is statistically significant for the  minor or non-bedridden illnesses.  The




results   also   demonstrate  that   the   additional   effort   involved   in




constructing the lifetime smoking profiles was worthwhile:   in  many of the




specifications of  the  respiratory illness measures,  both the  current rate




of consumption  and the  cumulative lifetime  consumption were  statistically




important  determinants  of  the  probability   of respiratory  illness.    In




addition, the last  section  of  Chapter 6  produces  some calculations of the




relative  risks  of  various  respiratory   illness  outcomes  attributable   to




plausible hypothetical changes  in ambient air pollution concentrations and




current and lifetime cigarette consumption.

-------
                                    16






     Chronic rather than acute  illness  is  the focus of vie seventh chapter




of Volume III.   In particular,  the focus  there  is  on the possible role  of




ozone,   sulfates,  and  total   suspended  particulate  matter  in  chronic




respiratory disease  (CRD).   The  analysis  in Chapter 7  improves upon the




analysis  of CRD  in  Volume I  in  several  ways.    First,  the  sample  of




individuals used was restricted to respondents who has been living in  their




present location for  ten  years  at the  time of the  1979 HIS (compared  with




five years in the original work).   Also, the more sophisticated measures  of




lifetime smoking described  in   Chapter  6 were used.   Third,  the sample  of




respondents  was divided  into  two groups:    one  group  was  comprised  of




individuals who received the  special  "probe" questions  in   the  1979 HIS




pertaining to respiratory disease; the other group consisted of individuals




who received one of the five other disease probes.   The sample was divided




in this way because of evidence of differential response rates.




     The results of this reanalysis of air  pollution and CRD are  ambiguous,




in much the same way  as were the  findings in Volume I.  For instance,  when




ozone was measured by the 1979  annual average daily maximum at the nearest



monitor, it was positively  and  significantly related to CRD in the "probe"




group.     When  averaged  over   more  than  one  monitor   in  an   area,  the




coefficient on  ozone  declined  somewhat  in  significance,  and  it was by any




reasonable  standard  indistinguishable from  zero where the multiyear  ozone




data were  used.   In  regressions  using respondents from  the  probe group,




neither  sulfates  nor  total  suspended   particulates  were   significantly




associated   with   CRD,   nor   were    any   of   the   other   independent




variables—including  the  smoking  measures.   In the  regressions run  using

-------
                                    17






the non-probe  group,  ozone  was  never  significantly  associated  with  CRD.




However,  when  the  multi-year  averaged  air  pollution  data  were  used,




particulates were  positively and  significantly associated with  CRD.    So,




too, were both current and prior smoking habits, as well as income.




     On  the whole,  then, reflecting  on  both the  original  work  and  the




recent  reanalysis,  no consistent relationship  between ozone  (or any other




air pollutants)  and  CRD  has emerged in  the RFF work.   It  is  an area  of




continuing investigation, however.




     Chapter 8 presents the results of a variety of sensitivity analyses  of




the Volume  I research.   That  is,  the  results discussed  in this chapter




summarize analyses of  the sensitivity of the Volume I empirical results  to




different assumptions about  various  data and  model  specification issues.




Specifically, Chapter 8  treats  the  following topics.   Discussed  first  is




the question  of  proper specification of the precipitation measures.    Our




analysis suggests that the measures used in  the models in both Volume T and




in  several  of the  chapters of Volume  III   are  appropriate.   Second,  we




assess  the  effects  of sample size and  selection  on a set of the estimates




presented in Volume I.  It is found that the  implications of  varying sample




size are  largely  manifested  in the efficiency  properties  of  the parameter




estimates rather  than in  inferences  about  the central tendencies  of  the




estimates themselves.




     The third section of Chapter 8 presents a reestimation  using Poisson




regression  of  some of the  models estimated by ordinary  least  squares  in




Volume  I.    The  upshot of this exercise  is that  the  alternative Poisson




estimation  techniques  appear  to  be  superior  on  statistical  grounds.

-------
                                    18






However,  the   inferences   about  the  relationships   between  ozone  and




respiratory health  drawn  from  the corresponding models  in  Volume  I are




largely corroborated by the reanalysis.   The  analysis  discussed in Section




8.U concerns  the question  of  structural  aggregation.   Specifically, this




section   examines   whether   structures  of   the   relationships   between




explanatory   variables   and   respiratory  health   outcomes   differ  for




individuals with  different cigarette  smoking behaviors  or  with different




chronic  illness  statuses.     The  results  generally  suggest  that  the




hypothesis of  structural  homogeneity  across  the  various  smoking status or




chronic illness categories be rejected.

-------
                                    TECHNICAL REPORT DATA
                             (Please read Instructions on the reverse before completing)
1  REPORT NO.
  EPA-450/5-85-005d
                               2.
                                                              3. RECIPIENT'S ACCESSION NO.
A TITLE AND SUBTITLE
                                                              5. REPORT DATE
  Executive Summary Ambient  Ozone and  Human Health:
  An Epidemiological Analysis
                                                                June  1985 (Date  of  Preparatio
              6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)

  Paul R. Portney and John  Mull any
                                                              8. PERFORMING ORGANIZATION REPORT NO
9 PERFORMING ORGANIZATION NAME AND ADDRESS
  Resources for the Future
  1616 P Street N.W.
  Washington,  DC 20036
              10 PROGRAM ELEMENT NO.
                12A2A
              11 CONTRACT/GRANT NO
                68-02-3583
12. SPONSORING AGENCY NAME AND ADDRESS
                                                              13 TYPE OF REPORT AND PERIOD COVERED
U.S. Environmental Protection Agency
Office  of  Air Quality  Planning and  Standards  (MD-12)
Research Triangle Park,  NC 27711
                                                              14.'S
                           G''£NCY CODE
                                                                OAQPS
15 SUPPLEMENTARY NOTES

  Project Officer:   Thomas G.  Walton
16 ABSTRACT
       This report is the executive summary of an analysis of the  relationship
   between ozone  and human health benefits.
17.
                                 KEY WORDS AND DOCUMENT ANALYSIS
                   DESCRIPTORS
                                                b.IDENTIFIERS/OPEN ENDED TERMS
                                                                               COSATI Held/Group
   Benefit Analysis
   Air Pollution,  03
   Epidemiology
18  DISTRIBUTION STATEMENT
   Release Unlimited
19 SECURITY CLASS (This Report I
 Unclassified
                                                                            21 NO OF PAGES
                                                                               21
                                               120 SCCUPITY CLASS i This page <
                                                  Unclassified
                                                                            22. PRICE
EPA Form 2220-1 sRev. 4-77)
                       = REVlOJS EDITION S OBSOLETE

-------
                                                          INSTRUCTIONS

  1.   REPORT NUMBER
       Insert the LP.A 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
       "itle should indicate clearly and briefly the subject coverage of the report, and be displa> ed prominently.  Set subtitle, if used, in smaller
         je or otherwise subordinate it to mam  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
       Farh 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 oj
        ,,,  oval, dare 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,I  List author's affiliation if it differs from the performing orgam
       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 hirearchv

  10.  PROGRAM ELEMENT NUMBER
       Use the program element number under vvhich 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 .it
       To be published in. Supersedes, Supplements, etc

  16.  ABSTRACT
       Include a brief 1200 words or less) factual summary of the  most significant information contained in the report If the report Contains i
       significant bibliography or literature  survey, mention it here.

  17.  KEY WORDS AND DOCUMENT ANALYSIS
       lal 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  I IELD GROUP - Field and group assignments are to be taker from the 1965 COSAT1  Subject Category List   Since the ma-
       jority of documents are multidisciplmary in nature, the Primary Field, Group assignmenUsI will be specific discipline, area of human
       endeavor, or type of physical object. The apphcation(s) will he cross-referenced with secondary Field/Group assignments that will follow
       the primary posting!s)

  18.  DISTRIBUTION STATEMENT
       Denote releasabilitv  to the public or limitation for reasons  other than security tor example "Release Unlimited." Cite anv  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 b> the National Technical Information  Service or the Government Printing Office, if known
EPA Fo—i 2220-1 (Rev.  4-77) (Reverse)

-------