You are on page 1of 16

Journal of Health Economics 28 (2009) 688–703

Contents lists available at ScienceDirect

Journal of Health Economics
journal homepage: www.elsevier.com/locate/econbase

Air pollution and infant health: Lessons from New Jersey夽
Janet Currie ∗ , Matthew Neidell, Johannes F. Schmieder
Columbia University, Department of Economics, International Affairs Building, 420 W. 118th Street, New York, NY 10027, United States

a r t i c l e i n f o a b s t r a c t

Article history: We examine the impact of three “criteria” air pollutants on infant health in New Jersey in the 1990s by
Received 22 July 2008 combining information about mother’s residential location from birth certificates with information from
Received in revised form 25 January 2009 air quality monitors. Our work offers three important innovations. First, we use the exact addresses of
Accepted 11 February 2009
mothers to select those closest to air monitors to improve the accuracy of air quality exposure. Second,
Available online 27 February 2009
we include maternal fixed effects to control for unobserved characteristics of mothers. Third, we examine
interactions of air pollution with smoking and other risk factors for poor infant health outcomes. We
JEL classification:
find consistently negative effects of exposure to carbon monoxide (CO), both during and after birth, with
I18
Q53
effects considerably larger for smokers and older mothers. Since automobiles are the main source of carbon
monoxide emissions, our results have important implications for regulation of automobile emissions.
Keywords: © 2009 Elsevier B.V. All rights reserved.
Air pollution
Infant health
Carbon monoxide
Birth weight
Infant mortality

The primary goal of pollution abatement is to protect human of a geographic area or computing average pollution levels within
health, but there is still much debate about the specific health the geographic area. In our data we know the exact addresses of
effects. This paper addresses this issue by examining the impact mothers, enabling us to improve on the assignment of pollution
of air pollution on infant health in New Jersey over the 1990s. Pol- exposure.
icy makers and the public are highly motivated to protect these Despite this improvement in pollution measurement, we must
most vulnerable members of society. There is increasing evidence still confront the problem that air pollution is not randomly
of long-term effects of poor infant health on future outcomes; for assigned, making potential confounding a major concern. Since air
example, low birth weight has been linked to future health prob- quality is capitalized into housing prices (Chay and Greenstone,
lems and lower educational attainment (see Currie (2008) for a 2003a,b) families with higher incomes or preferences for cleaner air
summary of this research). Studying infants also overcomes sev- are likely to sort into locations with better air quality, and failure to
eral empirical challenges because, unlike adult diseases that may account for this will lead to overestimates of the effects of pollution.
reflect pollution exposure that occurred many years ago, the link Alternatively, pollution levels are higher in urban areas where there
between cause and effect is more immediate. are often more educated individuals with better access to health
Our analysis improves upon much of the previous research by care, which can cause underestimates of the effects of pollution.
improving the assignment of pollution exposure from air quality Our data permits us to follow mothers over time, so we include
monitors to individuals. Most observational analyses that assess both pollution monitor and maternal fixed effects to capture all
the impact of air pollution on health assign exposure to pollution time-invariant characteristics of the neighborhood and mother. In
by either approximating the individual’s location as the centroid our richest specification, the effects of pollution are identified using
variation in pollution exposure between children in the same fam-
ilies, after controlling flexibly for time trends, seasonal patterns,
weather, pollution monitor locations, and several observed charac-
夽 We are grateful for funding under NIH grant R21 HD055613-01. All opinions teristics of the mother and child.
and any errors are our own. We would also like to thank Katherine Hempstead Infants at higher risk of poor outcomes may be differentially
and Matthew Weinberg of the New Jersey Department of Health for facilitating
affected by pollution, so we also examine whether pollution has a
our access to the data. Seminar participants at Tilburg University provided helpful
comments.
differential impact on infant health depending on maternal charac-
∗ Corresponding author. Tel.: +1 212 854 4520; fax: +1 212 854 8059. teristics, such as whether the mother smoked during pregnancy and
E-mail address: jc2663@columbia.edu (J. Currie). older maternal age. Previous research has suggested that smoking

0167-6296/$ – see front matter © 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.jhealeco.2009.02.001

ing countries that are likely to have comparable levels of pollutions to New Jersey For example. Woodruff et al. Ritz et al. These findings for CO are robust between very severe pollution episodes and increased mortality to many different specifications. 2005. limiting our review to develop- our results. (2000) report that PM10 exposure 6 occurs are not well understood. Wilhelm and Ritz. Ritz in cigarette smoke (World Health Organisation.. Bell et al.. al. and sensitizes the lungs to other irri- lutants. Parker et al. reducing the infants with high exposure to PM10 are more likely to die in the body’s ability to deliver oxygen to organs and tissues. Liu et al. Chen et al.. 2006. Salam et al. 2003).. the placenta and accumulate in the blood of the fetus. (which are found in auto emissions. J. these pollutants may themselves be markers for other com- weakens the immune system (Seaton et al. the first 2 weeks after birth increases the risk of infant mortality Many epidemiological studies have demonstrated links by 2. 2000. is correlated with gestation in some regions of southern California pected although the exact biological mechanisms through which it but not in others. we estimate that a one unit change in mean CO impacts. 2008. For example. the availability of oxygen to be transported to the fetus. Moreover. Section 1 provides Previous epidemiological research on the effects of moderate necessary background about the ways in which pollution may affect pollution levels on prenatal health suggest negative effects but have infant health and the previous literature.. al. acetonitrile. 2000). they would have to damage the fetus aromatic hydrocarbons (PAHs). 1998).gov/ttn/atw/hapindex. developing fetuses in animal studies. In pregnant women. Chart 1 have and motor vehicle exhaust is a major source. infants are particu.. among other sources) in heat matory responses and airway reactivity (Xu and Wang. (1997) report that CO bonds with hemoglobin more easily than oxygen.. Many of the studies in Basu et cleared from fetal blood more slowly than from maternal blood. Studies in humans Ozone (the major component of smog) is formed through reac. Sagiv et Particulate matter can take many forms. in smokers than in non-smokers if the effects of pollutants are coughing. for example. For example. with a lead.. and cyanide (see http://www. much of the negative effect of Lee et al.1 Although the available research points towards potential birth. An important limitation of these studies is that the observed carbon monoxide readily crosses the placenta and binds to fetal relationships could reflect unobserved factors correlated with both haemoglobin more readily than to maternal haemoglobin and is air pollution and child outcomes. 2008. the results are not respiratory problems. such as retarded growth. colorless gas that primarily comes from transportation sources. reduces lung function. including carbon monoxide (CO). But Lipfert et al. 2007. (2001) find that PM10 has no effect on low birth weight. 1953). including ash and dust. U. all three pollutants can directly affect infants after effects on infant mortality conditional on measures of health at birth. It has been areas is automobile exhaust. Section 4 presents studies examining this relationship. today have effects on infant health. leading to concentrations that may be 10–15% higher in the fetus’s Dugandzic et al. 2003. tissue. Carbon monoxide is an odorless. 2006. 1997). and pulmonary congestion.. since cigarette smoke contains high levels of pol. Furthermore. Since the major source of CO in urban cardiopulmonary mortality (Logan and Glasg.. mixed results. 2008. 2005. Indeed. Friedman et al.. this is the larger effects than other pollutants because of its ability to cross first study to ask whether there are such differential effects.. weakening her immune sys- experienced in New Jersey in recent years. (2000) report that CO exposure in the 6 weeks before birth A link between air pollution and infant health has long been sus. Parker et al. infants may already be tants. have shown elevated levels of an enzyme induced by PAHs in women about to have tions between nitrogen oxides and volatile organic compounds preterm deliveries (Huel et al.. 2006. exposure to ozone during exercise reduces lung exposed to high levels so that the marginal impact may be smaller functioning in adults and causes symptoms such as chest pain. It is not clear why ozone non-linear. We also find that the effects of CO of infants and others. Moreover. with as much as 90% of CO in cities coming from motor vehicle Studies of the effects of pollution on infant mortality also yield exhaust (Environmental Protection Agency. very minimal (if any) controls for potential confounders. since motor vehicle exhaust is a major contributor of CO ing theory being that they cause an inflammatory response that and PM10.. 2005. 2006. and sunlight.. though like PM10 it might indirectly affect mothers might be more susceptible to problems related to smoking the infant by compromising the mother’s health. Ritz and Yu (1999) report that CO expo- sure in the last trimester of pregnancy increased the incidence of 1. One of the most famous focused on a on infant health at birth are two to six times larger for smokers “killer fog” in London. Families ticles are widely believed to cause the most damage since they with higher incomes or greater preferences for cleaner air may be are inhaled deep into the lungs and can possibly enter the blood- stream (Environmental Protection Agency. Ritz et al. while data are described in Section 3. exposure to CO reduces robust to including controls for maternal characteristics. weeks before birth increases preterm birth. 1993). Rogers and Dunlop. In particular. Our estimates confirm that carbon monoxide has a significant pollution exposure could indirectly affect the fetus through the effect on fetal health even at the relatively low levels of pollution health of the mother by. 2007. 2002.. it provides little guidance about the necessary levels of during the last trimester of pregnancy increases the risk of low birth pollution to induce negative effects or when fetuses or infants are weight by 8%. Currie et al. 2004. England and found dramatic increases in and for mothers over age 35. these findings have implications for less clear whether levels of air pollution that are common in the regulations of automobile emissions. and that it has further tem. While CO post neonatal period. / Journal of Health Economics 28 (2009) 688–703 689 might exacerbate the effect of air pollution by increasing inflam. 1995). smoking on infant health is believed to be due to the CO contained Parker and Woodruff.. butadiene. The mechanisms through which particles harm health are controversial. Background low birth weight (defined as birth weight less than 2500 g).epa. Since particles ponents of exhaust which injure infants. Brauer et al. 2005. and Section 5 details our conclusions. Huynh et al. (2000) find that although is poisonous to healthy adults at high levels. 2008... 2003). Section 2 describes our produced inconsistent results. Liu et al. while Maisonet et al. benzene. Rogers et al. 2001. January 1993. To our knowledge. (Cnattingius. blood than in the mother’s. they can reproduce some earlier results showing effects of county- larly susceptible because they are smaller and often have existing level pollution measures on infant mortality. 2007. The smallest par. . et al.5% relative to baseline levels. However.S. so it is also possible that these infants are more The discussion suggests that one might well expect CO to have vulnerable to the effects of pollution. The rest of the paper is laid out as follows.html) have been shown to have effects on indirectly by provoking inflammation in the mother. a one unit change in mean CO during most vulnerable. Previous work has also suggested that infants of older would affect the fetus. 1 Alternatively. while Ritz et al. Components such as polycyclic cannot cross the placenta. Chart 1 provides a list of previous methods. Ozone is a highly reactive compound that damages Alternatively.

developed countries. 690 J. Currie et al. . Selected epidemiological studies of effects of pollution on infant health. / Journal of Health Economics 28 (2009) 688–703 Chart 1.

So for a given individual much of the variation in air quality we also estimated models using the maximum daily value of pollution over the same comes from variation in ambient pollution levels. for example. pollution levels (Samet et al. One might expect. The vector Pmt contains measures of ambient pollution levels births. In order to control for Actual personal exposures are affected by ambient air quality. they were teristics include dummy variables for the mother’s age (19–24. 5 indoor sources of air pollution may be relatively constant and uncorrelated with While this measure captures high ambient levels sustained over a period of time.2 and . an indicator for whether it using more accurate measures of pollution exposure. They also find some evidence that the decline in TSPs led in each of the first. denoted by s. only TSPs from a change in mean pollution levels for trimester s. and by examining other pollutants in addition to particulates. suggesting the overall direction of bias from confounding In order to examine the effect of pollution on health at birth. They estimate that a one unit decline in particulates caused by the where O is a birth outcome. 35+). but found that it was not statistically significant in any of our models.1. Hence. Given that family income rier of the placenta or because it has a systemic effect on the is not included on the birth certificate. Research on the relationship between indoor and outdoor air quality (Spengler experience less economic activity in winter than in summer relative et al.000 live ods. or 16+ years).2 Two studies by Chay and Greenstone (2003a. to preserve sample size we control for missing values by including an additional “missing” category for As discussed in the previous section. An additional issue is that this paper (like the others discussed above) tion exposure is likely to be correlated with omitted characteristics examines the effect of outdoor air quality measured using monitor in fixed locations. mother’s education (12. Before birth. whether the mother is married. We construct the trimester measures by taking the aver- higher than those prevalent today. and individuals in urban areas may be more educated and have better access to health care. such as improvements in health examine the effect of changes in traffic patterns in Atlanta due to the 1996 Olympic care. games.. 2000.6 in most studies of PM10 4 for e. After birth. who Yt includes month and year dummy variables to capture seasonal investigated the health effects of the temporary closing of a Utah steel mill (Pope.4 so ˇs reflects the effect fallen by nearly 50% from 1980 to 2000. they find that a one istics taken from the birth certificate that are widely believed unit reduction in carbon monoxide over the 1990s in California to be significant determinants of birth outcomes. indoor air quality. a limitation of model (1) is that pollu- dell (2005). / Journal of Health Economics 28 (2009) 688–703 691 more likely to sort into neighborhoods with better air quality. 1992.3 and other or unknown race. 3 Smoking data was not available in the California data used by Currie and Nei- As previously mentioned. we examine the If these factors are unaccounted for.b) deal with the miles) of a monitor and estimate baseline models of the following problem of omitted confounders by focusing on “natural experi- form: ments” provided by the implementation of the Clean Air Act of 1970 and the recession of the early 1980s. i indexes the individual. 2000) suggests that much of what is outdoors comes indoors. represents daily precipitation and daily minimum and maximum inates their ability to examine other pollutants that are correlated temperature averaged over each trimester of the pregnancy. we also include a measure of median family income and the fraction of poor households in 1989 in the mother’s census block group as a proxy.. dummy variables for missing controls. whether the mother is African-American. Since these vari- 2. whether the for mother fixed effects. control for weather in the vector w because it may have inde- Currie and Neidell (2005) extend this line of research by exam. the time-series correlation is higher. We with particulates emissions. controlling is a multiple birth. and t indexes time peri- and eight (four and seven) fewer infant deaths per 100. this would lead to an upward effect of pollution on infant mortality conditional on health at bias in estimates. second. pollution with smoking and other risk factors. Wilson et al. . restrict the sample to women who lived within 10 km (about 6. However. m indexes the nearest monitor. infants are directly exposed to families are also likely to provide other investments in their chil. 3rd. effects (pollution is strongly seasonal and birth outcomes may also 1989. sured by birth weight and gestation. pendent effects on birth outcomes and is correlated with ambient ining the effect of more recent levels of pollution on infant health. Pope et al. j indexes the implementation of the Clean Air Act (recession) led to between five mother.000 live births. Appendix Table 1 shows the complete specifica- infants differently before and after birth. 13–15. and investigating the interaction of air child is male. However. Hispanic. Methods ables are all categorical. unable to find any consistent evidence of pollution effects on health 25–34. This is because for a given individual We describe these trimester measures in more detail in the following section. These charac- saved 18 infant lives per 100. sonal effects in these characteristics (for example. and third trimesters of the mother’s to reductions in the incidence of low birth weight. PM10 levels have age pollution measure over the trimester. Alternatively..2 Both the Clean Air Act and  3 s the recession induced sharper reductions in particulates in some Oijmt = (Pmt ˇs + wmt s  s ) + xijmt ı + Yt + εijmt (1) counties than in others.). inhaled pollutants. although the cross-sectional correlation between ambient air quality and personal exposure is low (between . and at birth. 4th or higher). coastal areas that infants spend little time outdoors so that outdoor air quality might not be rele- vant. so that fetuses and infants exposed to lower levels of pollution we examine the effects of pollution on health at birth as mea- also receive more family inputs. Furthermore. and omitted characteristics of neighborhoods and for differential sea- the time the individual spends indoors and outdoors. J. using the monitor closest to the mother’s levels of particulates studied by Chay and Greenstone are much residence. (2001) who be) as well as trends over time. Modeling birth outcomes may improve health. Furthermore.. Currie et al. Using The vector xijmt includes mother and child specific character- within-zip code variation in pollution levels. and they use this exogenous variation in s=1 levels of pollution at the county-year level to identify its effects. Second.g. these studies did not look specifically at infants. our analysis proceeds in two parts: First dren. This paper improves on Currie and Neidell (2005) by birth order (2nd. we is unclear. be located in urban areas. and whether the mother smokes. intervals. 1992) and to Friedman et al. 1997). of families that are related to infant health. such as better quality prenatal care. However. Arden Pope. Ransom and Pope. These health of the mother.5 The wmt were measured during the time period they examine. outdoor air quality. the pregnancy. factors that 2. and the number of cigarettes if she smokes. Omitting these factors would lead to a down- ward bias. The vector 2 These studies are similar in spirit to a sequence of papers by C. pollution emission sources tend to birth. air pollution may affect each variable. pollution tion for one of our models that includes the coefficients on the may affect infants either because it crosses the protective bar. which elim. for example.

and s Oijmt = (Pmt ˇs + wmt s  s ) + xijmt ı + Yt + ϕmt ∗ Qt + ςj + εijmt (3)  4 for over 6 weeks. unmarried (62% vs. Allison (1982) shows that estimates from models of infant health. and 0 in all other periods.7 standard deviations for ozone are 0. We allow for such differential effects by interacting the weekly richest specification we estimate: pollution measure Pmt with 4 dummy variables   indicating time since birth. we are concerned about the impacts of pollution on the infant between births. birth. insensitive to including birth weight as a continuous covariate (such as birth parity) is repeated for every period. We choose the same quarter to capture average neighborhood characteristics break points after 1. at the life/death margin.2.S. we modify the birth outcomes model to capture the fact that birth outcomes are a one-time occurrence but mortality is a contin. These models control for as well as by pollution after birth. Although the average and 0. family variation would have been if each mother had stayed in the 72% overall) and who are smokers (13% vs.271. 20. Each time-invariant 6 Our results are. for ozone.5% overall).939 and 0.   between 2 and 4 weeks.6 To the extent that birth weight is a sufficient statistic for due to air quality regulations. then our results will tell us about the effects of variations In order to examine infant mortality conditional on health at in pollution for the range of pollution we observe. Since the infant death hazard varies ular. were more likely to be low birth weight and mothers systemati- This procedure yields a very large number of observations since cally moved to cleaner environments between the first and second most infants survive all 52 weeks of their first year. / Journal of Health Economics 28 (2009) 688–703 to inland areas). we will instead identify a local updated each period. 4. Environmental Protection who lived for n weeks as if they contributed n person-week obser- vations to the sample.  3 between 4 and 6 weeks. 1500–2500 g. respectively. We then estimated what the within who are African American (30% vs. pollutants in a given week. and within the year. 3. This suggests that mothers do family with a death is different than the average family without not appear to be systematically moving to cleaner or dirtier areas one. 19% overall). If the characteristics of the marginal infant who dies because of an increase in pollution is similar to the char- 2. families with a death are more likely to have mothers ily” variation in each pollutant. may be important for her infant’s health and may on infant death. and for PM10 are 0. Therefore. specified as a lin- of children who live in close proximity to each other and are born in ear spline function in the weeks since the infant’s birth.  1 equals one if time since birth is between 0 and 2  3 weeks. due to sea- health at birth. 2500–3500 g. 3500 g). the time-varying covariates (such as pollution and weather) are 7 To the extent these conditions are not met. location in which she was first observed.410 heterogeneity in the vulnerability of infants. we performed to mothers who lost at least one child.301 and 0. we estimate a model in which the probability of death Dijmt is specified as  3 Oijmt = s (Pmt ˇs + wmt s  s ) + xijmt ı + Yt + ϕmt ∗ Qt + εijmt (2)  4 s=1 Dijmt = ˛(t) + (  Pmt ˇ + wmt s  s ) + xijmt ı =1 where now ϕmt is a fixed effect for the closest air pollution monitor +Oijmt + Yt + ϕmt ∗ Qt + ςj + εijmt (4) and ϕmt *Qt is an interaction between the monitor effect and the quarter of the year. One way to think about these estimates is in terms of underlying for CO are 0. However. It would be a problem. average treatment effect. Because infant death might be affected by pollution before birth where  j is a mother-specific fixed effect. Hence. The within family vari- mean ozone. in our ard. In terms of observable char- the following exercise. haz- A necessary condition to identify the impact of pollution is that ard model that allows for time-varying covariates. Data uously updated outcome. This model can be thought of as a flexible. CO. 9. if first children this type converge to those obtained from continuous time models. so that the effects of pollution are identified by variation in variables. would also vary with the baseline haz- also be correlated with her choice of neighborhoods. Currie et al. we limit this part of the analysis variation in pollution is uncorrelated with mobility. . To do this. accessed from the technology transfer network air quality tion over the infants’ first year of life. Much of of dummy variables (<1500 g. and 32 weeks to capture the shape within a season. In order to check that the reduce the number of observations. if it exists. In order to births because their incomes increased. We control for birth weight flexibly by including a series pollution at a particular monitor between pregnancies. respectively. activity. while variable. Thus the effect of pollution as measured by ˇ s=1 can differ arbitrarily over these four intervals.692 J. 12. itoring. we treat an infant system database maintained by the U. (4) will capture the independent effect sonal patterns in pollution and unpredictable variations in human of pollution after birth conditional on health at birth. Pmt measures exposure to the three Model (2) may still suffer from omitted variables bias. 2. 8.947. unobserved characteristics of mothers. however. In this specification. we compare the outcomes where ˛(t) is a measure of duration dependence. it is likely that an effect of pollution her own health. The dependent variable is coded as 1 in the period the infant dies.407. and PM10 measures in the trimester before birth ances were virtually identical: the actual and simulated within are virtually identical in families with deaths and those without. non-parametric variation in infants’ pollution exposure is uncorrelated with other duration dependence. of the actual empirical hazard. respectively. Detailed data on atmospheric pollution come from the New we estimate a weekly hazard model with time-varying covariates Jersey Department of environmental protection Bureau of Air Mon- to account for a varying probability of survival and levels of pollu. the risk of death is highest in the first week or two of life and drops sharply thereafter. such as her regard for greatly with time since birth. Model for infant mortality acteristics of the marginal infant who survives the same increase in pollution. we add birth weight as a measure time-invariant characteristics of both neighborhoods and moth- of infant health outcomes at birth (Oijmt ) to the list of independent ers. ˇ from Eq. For example. monitor-specific quarter effects and mother characteristics of the infant or the infant’s families that may affect fixed effects. and over this variation is driven by changes in pollution levels over time. In partic. We first estimated the actual “within fam- acteristics. for example. we estimate models of the form: Based on this data structure. discrete-time.

These correlations are even higher within monitor. Although we choose these measures because they standards are based on daily maximum concentrations. 1. the ozone are 0. and 9. Currie et al. the EPA calculates 8 h moving average values. the average within a 24 h period. we construct measures of pollution by taking levels of air quality as mothers with a single birth. J. without.11 Of the 1. the 3-year moving average of the fourth-highest daily CO and daily mean for CO is 0. subsets of the geocoded sample. New Jersey to use a confidential version of the data with the mother’s address.noaa. For example. 9 12 The 8-hour maximum corresponds to taking the maximum 8-period moving Although these mean pollution levels are far below air quality standards.ncdc. name. with the sample becoming much smaller but still sizable at 21. we interpolated using data from surrounding counties. 1 demonstrates that monitors are heav. Since PM10 is not measured every day. Finally. / Journal of Health Economics 28 (2009) 688–703 693 Surface Summary of the Day (TD3200) from the National Climatic Data Center. 24-h average of PM10.7 in the sample closer tion than the rule for a monitor location to measure all three of the to monitors. and control variables are shown in Table 1. where smokers and nonsmokers live. For the pollutants of interest.epa.9 per 1000 births overall. education. such as race. and 4 of Panel A shows that infant health is worse in the population closer to mon- Agency (EPA). For . If weather data was not available for a county and date. maternal characteristics between successfully and unsuccessfully geocoded mothers. though we will also show one specification that sibling regression models.91 and 0. however. which lie suggests no systematic difference in air quality between the areas along the transportation corridor between New York and Philadel. 36% were successfully geocoded and within 10 km of an ozone monitor. Because different monitors measure different pollutants. A comparison of columns 1 and 2 shows no differences in Fig. followed by O3 and CO.10 Data on infant births and deaths come from the New Jersey Department of Health birth and infant death files for 1989 to 2003. the one measures is shown in Fig. A comparison of columns 2. Similarly. Vital Statistics records are a very rich source of data that cover all births and deaths in New Jersey. that infants pollutant.dll?wwAW ∼MP#MR. We traveled to Trenton. 1. the sub- samples used in the regression models are slightly different.gov/cgi-win/wwcgi.gov/ttn/airs/airsaqs/detaildata/ Sample sizes also vary slightly for different outcomes because of missing values downloadaqsdata. which is the final sample we use in our analysis. 3.93. For determining compliance are based on air quality standards. mothers with more phia. which improves our ability to generalize results from the els (and samples).8 The location of each of 57 monitors and what each itors. respectively.12 the mean of the daily values either over the three trimesters before birth (for the birth outcomes models) or for each week after birth (for the infant mortality model). includes both CO and O3.94. air quality standards. Unfortunately. and our daily measures we use are the 8-h maximums of CO and O3 and the models incorporate monitor fixed effects. and the sample of births to smoking mothers within 10 km of an ozone monitor. Location of air monitors in New Jersey. and marital status.htm>. PM10 is the most frequently monitored column 5 or column 4 to column 6 suggests. the sample of births within 10 km of an ozone monitor. death rate is 6. for the outcomes. we will examine the impact of each pollutant in separate mod. than one birth over the sample period are exposed to comparable For each monitor. the correlation then asks whether the daily maximum of this moving average ever exceeds 9 ppm between the maximum 8 hour reading for CO with the maximum 1 hour average for during the year. the two pollutants that have the largest Panels B and C give means of the pollution measures for the effects individually. Birth records have both detailed information about health at birth and background information about the mother. For ozone. For example. Column 5 restricts the sample further to children with a sibling within the sample. Almost 20% of the total births are in the sibling sample and within 10 km of a monitor. A comparison of columns 3 and 4 ily clustered in the most populated areas of the state. Fig.08 ppm. Births were linked to the air pollution measures taken from the closest monitor by using the mother’s exact address and the latitude and longitude of the monitors. the sample of births with residential address that were successfully geocoded. pollution mea- sures. column 6 further restricts the final sample to the subset of mothers who smoked at both births. Comparable correlations for maximum 8-hour average ozone concentrations must be less than . Comparing column 3 to pollutants that we study. Because of this limitation of the with siblings in the sample do not differ systematically from those data. and common measures of short-term spikes in pollutants. the measures are highly correlated with other with air quality standards for CO. It was also possible to link birth and death records to identify infants who died in the first year of life. 7.9 among the smokers. and much worse in the sample of smokers.75 million births in New Jersey over our sample period.9 County level weather data come from the 10 This data is available at http://www4.98 and 0.099 births. The use of this data allows us to more precisely match mothers to pollution mon- itors and to identify siblings born to the same mother. it is more the excep. and birth date. 8 11 The data is available at: <http://www. Our tests of this procedure (using counties with weather data) indicated that it was highly accurate. The first four columns show means for all births in New Jersey. Descriptive statistics for infant health outcomes. which correspond with national ambient the weekly mean for PM10 may be noisier than those for other pollutants. with roughly 10% of these births to mothers who smoked.

4 2937.694 J.epa.029 0.129 0.34 22.6 3236.732] PM10 (24 h moving average in 10 ␮g/m3 ) 2.992] [6. Currie et al.164] [5.035 1.51 0.861 1.584] [1. but there is a substantial amount of variation remaining to identify health period for these graphs (1994 to 1998) is restricted to improve exposition.72 29. fixed effects and mother fixed effects reduces the standard deviation to 0.528] CO (8 h moving average in ppm) 1.35 13.74 11.143] Mean precipitation in previous 13.19 0.58 1.42 64.643] [3.59 1.64 1.000) [1.99 2.157 0.44 3.615] [1. there were 2 ozone monitors in 0.01 ppm) 3.822] [1.79 11.212] Low birth weight 0. several CO monitors experienced AQS violations in the period (e.67 previous 90 days [14.600] [2.475] [2.145] [1.87 22.61 0.html).938] [5.0086 0.57 [1. and the weather data.g.05 3.621] [1.00 0.83 38.27 12.51 0.211] [4.97 3.032 0. Figs.375] [1.97 3.805] [1.35 [3. respectively.507] [1.5.995] [6.120] [0.14 Panel D of Table 1 shows means monitor (the Camden Lab monitor in Camden) over time and resid.26 21.95 2. while the “b” 14 While these figures are on the monitor level.393 21.14 0.54 Mother Hispanic 0.72 0.22 28.25 tract 1989 ($10.003] [5.4 [617.51 [0.20 [0.099 Panel A: outcomes Birth weight in grams 3320.0073 0. mean CO levels started at 4 ppm at show that although adjusting for these factors accounts for sea- the beginning of the sample period.18] [15.55 1.746] [0. After taking average over three years (see http://www.28 38.55 [1.791] [1.672] [2.748] Panel C: pollution measures 1 week after birth Ozone (8 h moving average in 0.766] [1.92 [5.996 283.16 10.60 3.492] [1. As a group the controls account for a significant part of the varia- violation (1995 and 1998).946] [2.03 90 days [4.65] [14.55 37.10 1.33 [1.0078 0.74] Mean of daily min temperature 21.186] [1.172 0.060 0. this is reduced to 0.076 0.498] [1.00 Number of cigarettes per day 1.8 3267.04 22.21 and 4 out of 13 monitors in 1989) but none after 1995.565] [1. It is also important to note that the means in Table 1 mask con.15] [15.080] [4.29 Birth parity 1. effects.903] [3. In the most polluted areas.30 0.800] [1. there is still considerable variation left to by 2005.74 3.825] CO (8 h moving average in ppm) 1.14 0.703] [0.107 0.60 3.0099 0.09 0.2] Infant death 0.210 Panel B: pollution measures last trimester before birth Ozone (8 h moving average in 0.3 3054.495] [1.892] [2. / Journal of Health Economics 28 (2009) 688–703 Table 1 Sample means.11 12.60 3.70] [14.817] PM10 (24 h moving average in 10 ␮g/m3) 2.739] [0.46 [2.71 38.491] [1. The time dummies.105] [7.01 ppm) 3.0069 0.43 previous 90 days [15. tion in pollution.6] [682.17.99 [1.149] [4.503] [1. Taking out monitor * quarter of our sample.73 3. of the control variables available in the Vital Statistics data. 2–4 plot pollution levels at one particular pollution identify the effects of pollution. For the period out the controls in equation (1).069 Mother married 0.074] Mean of daily max temperature 63. and no PM10 monitors in violation.307] Fraction poor in census tract 1989 0.86 38.61 3.03 10. we also checked how much of the variation in pollution is absorbed by our regression controls on the mother level.25 27.66 4.13 0.23 0.53 3.881] [0.35 0.187 0.62 3.772] [0.158] [4.480] [1.502.64 1.98 2.129] [0.2 3319.01 [0.512 0. mostly because of the inclusion of seasonal controls and monitor 13 The patterns.06 1.625] [4.52 0.46 2.4] [630.74] [14.4] [615.00 2. and the weather variables included in our regression models.1] [660.96 2. the 24 h average must not exceed 150 ␮g/m3 more than once per year on For example for CO the standard deviation is 0. These plots time.758] [0.10 Mother years of education 13.44 27. PM10.84 [2.302] [2.97 2.856] Child male 0.632] [2.754. the ual pollution levels after controlling for the time and monitor effects decennial census.60 1.51 [0. not shown here.01 1.18] [15.99 2.17 0.74] [14.41 0.97 2.98 13.60 3. Column [6] contains births where the mother smoked during the pregnancy for at least one sibling.0128 Gestation 38.971] [3.792] [0.55 1.103] [0.744] [0.12 1.862] [0. but declined to roughly 1 ppm sonal and annual trends.524] [1. are very similar for the other monitors.20 0.70 64.10 64.205 628. .725 0.7 in the full sample.137] [0.51 Mother smoking 0.77 12.98 2.12] Notes: Standard deviations in brackets.51 0.0338 0.681] [1.05 13.571] Median family income census 4. [1] All [2] Geocoded [3] <10 km [4] <10 km monitor [5] Like (3) but [6] Like (4) but monitor and smoking ≥1 sibling ≥1 sibling Number of observations 1.6] [656.843] Multiple birth 0.737] [0.645] Mother African American 0.36 0.09 64.gov/air/criteria.288] [1.75 26.11] [15.911] [7.848] [0.59 0.148] [1.874 61. series plot 7 day moving averages (corresponding to the measures siderable variation in pollution levels both across monitors and over of pollution we use in the infant mortality models).077 0.956 1.026 0.18 0.09 0.089 0.60 1.796] [0.504] Panel D: control variables Mother age in years 28.02 13.13 The “a” series plot 3 month moving averages (corresponding to the mea- sures of pollution we use in birth outcome models).

are much less likely to be married. comparable levels of pollution. It is striking that mothers within 10 km of a monitor are also much more likely to be African Amer. Eq. as we do in our specifications. estimates of Eqs (1)–(3) for a different pollutant. J. Currie et al. wrong-signed. 2. mothers are typically exposed to higher levels of pollution. the than further census tracts. To assure that the differences between the tors are lower income and have a higher fraction of poor inhabitants models are not driven by changes in the sample composition. They are also less likely to be mar. Estimates of the effects of pollution on all mothers within 10 km ican or Hispanic and have half a year less education on average of a monitor are shown in Table 2. although urban panic). These patterns are consistent with resi. Failure to control for these factors could ers with more than one birth in the sample look quite similar to yield estimated coefficients that are biased down and possibly even mothers observed to have had only one birth. Furthermore. (3). In contrast. / Journal of Health Economics 28 (2009) 688–703 695 Fig. For example. Results on average than the sample mean. 2 children in the sample. ter access to health care. In all models we cluster standard errors at the cen- to the monitors are generally worse than those farther from the sus tract level to allow for common shocks to mother’s exposed to monitors. is only identified from mothers with at least further away from monitors. The mother fixed ried. 90 day moving average of CO. (4). (a) Air quality at Camden lab monitor. characteristics can be misleading. and when they are. but only slightly more likely to smoke than mothers who live effects model. negative ones. sample for estimating all three equations is restricted to children dential sorting based on air quality: monitors are generally located with at least one sibling in the sample (corresponding to column in more polluted areas. Each group of 3 columns shows compared to the full sample. Few of the pollution measures in columns (1). These systematic differences demonstrate the importance of and (7) are statistically significant. . census tracts near moni. and the characteristics of those closer (5) of Table 1). (b) Air quality at Camden lab monitor. they are as adequately controlling for characteristics of neighborhoods and likely to suggest positive effects on birth weight and gestation as families. 7 day moving average of CO. moth. Mothers within 10 km of a monitor are almost a year younger 4. The pattern of relative disadvantage is even more pronounced Table 1 suggests that the models that do not adequately con- for the population of mothers who smoke. These mothers are much trol for characteristics of the mother’s location and for her own more likely to be African-American (though less likely to be His. have a year less education. they are and live in the poorest census tracts compared to non-smoking also wealthier and more educated in our data and may have bet- mothers who live within 10 km of a monitor.

55 weeks)—a reduction in mean gestation neighborhood fixed effects to account for confounding factors. gests that infants at risk of low birth weight are most likely to be tive effect at the 10% level on birth weight and gestation. For PM10 affected by pollution.64 and standard deviation is 0. once we include monitor*quarter fixed effects (as in increase in the mean level of CO during the last trimester (where columns (2). However.65 g (from a base of 3236 g)—a reduc- probability of low birth weight. for a mother fixed effects. with maternal fixed effects. are very similar but are not shown). It of 0. smoking is estimated to have extremely neg- the estimated effect gradually increasing as we control more thor. and shortens gestation. mothers in more polluted areas have them to the effects of smoking. and (9). 7 day moving average of OZ. the estimates for CO become even larger. being a smoker is estimated to reduce birth of PM10 and ozone are inconsistent across specifications. The coefficients on smoking and unobserved characteristics that make them more likely to have number of cigarettes from the models for CO are shown in Table 3 healthy infants. lead to an increase in low birth weight of 0. The estimates in Table 2 imply that a one unit total reduction of approximately 212 g at the mean of 10 cigarettes .2%. increases the average birth weight by 16. unit change in mean CO is estimated to reduce gestation by 0. / Journal of Health Economics 28 (2009) 688–703 Fig. consistent with much of the prior oughly for potential confounders. (5). Additionally. a one significant at the 10% level. (a) Air quality at Camden lab monitor. when we control for mother fixed effects in columns 0. (the estimated effects of smoking in models for other pollutants To summarize: third trimester CO has statistically significant.0083 (from a base of Finally. the estimated effects literature. The proportional effects are greater only wrong-signed coefficient suggests that increases in PM10 in for low birth weight where a one unit change in mean CO would the first trimester of pregnancy increase gestation. with none weight by 162 g in models that include monitor fixed effects. an observation that we explore further below the first trimester in the low birth weight regression is statistically by examining infants with various risk factors. 90 day moving average of OZ. For example. Currie et al.074 tions suggests the importance of controlling for both maternal and week (from a base of 38. and statistically significant at the 95% level in the models that control for each additional cigarette smoked reduces birth weight by 5 g.79) would reduce last trimester of the pregnancy reduces birth weight. (b) Air quality at Camden lab monitor. The (3). (6).106)—an 8% increase in the incidence of low birth weight. 3. also suggests that in New Jersey.696 J. Now the tion of about a half a percent. Ozone greater effect for low birth weight than for mean birth weight sug- in the second trimester now has a statistically significant nega. and (8)) the estimates suggest that CO in the the mean is 1. ative effects on infant health. conditional on other observable One way to put these estimates into perspective is to compare characteristics of mothers. This pattern of results across specifica. In contrast. In models that do not include negative effects on infant health in all of our specifications.

Only the esti- the effects of CO are no longer statistically significant in the model mates on these interactions are shown. Similarly.2 in the model with estimates that apply to the rest of the sample) are generally compa- . 4. However.9 g.044 com- characteristics of the mother that are associated with her smoking pared to 0. and each cigarette smoking mothers. which shows estimates for mothers who smoked dur. per day. Furthermore. column 1 of Table 5 is based on the same ing both pregnancies. We investigate this issue tor of pollution measures with the relevant characteristic of the in Table 4. some demographic groups are fairly small. while PM10 in the first and second trimesters are both a roughly 3. (a) Air quality at Camden lab monitor. PM10 in the lent impact on birth weight as that from smoking 10 cigarettes second trimester is also estimated to reduce gestation significantly. differential effects were As discussed above. CO is −43 compared to −074 in Table 2. reduces the estimated effects of smoking mothers than for non-smoking mothers. These estimates indicate Including mother fixed effects. estimated to increase the incidence of low birth weight. Hence it would take birth weight. For gestation. 7 day moving average of PM10. Similarly. we now also find that PM10 in the reduces it a further 2. the effect of smoking 10 cigarettes per day is Table 5 places the results for smoking mothers in context by a bit more than twice as large as the impact of a one unit change showing estimates of the differential effects of CO on other subsets in mean CO in terms of the effect on the incidence of low birth of mothers who may be vulnerable to poor birth outcomes. though they remain large: being a smoker outcome. the point estimate of −39. measures are also interacted with an indicator for whether the lution exposure is more harmful to the infants of smokers. depending on the smoking considerably. Although mother was 19 years or younger at the time of birth. The point out. which controls for unobserved that the harmful effects from CO are two to six times greater for characteristics of the mother. J. these estimates are likely to be contaminated by omitted coefficient on CO in the models of low birth weight is 0. The point estimates in Table 4 are generally regression as column 3 in Table 2 except that the three pollution much larger than those in Table 2.2 g for a total reduction of about 61 g in infants second and third trimesters has a statistically significant impact on of women who smoke 10 cigarettes per day. 90 day moving average of PM10. (2005) and Tominey (2007) mother fixed effects is twice as large as the Table 2 coefficient. as the “main effects” (the for birth weight. per day. suggesting the same level of pol. (b) Air quality at Camden lab monitor.7 unit change in mean CO levels to have an equiva.008 in Table 2. the Table 4 coefficient on behavior. infants of smoking mothers could be either estimated using the full sample of births and interacting the vec- more or less affected than other infants. Since weight. the impact of ozone is four to six times larger for is estimated to reduce birth weight by 38. as Almond et al. mother. Currie et al. For example. / Journal of Health Economics 28 (2009) 688–703 697 Fig.

01 0.05 −2. Mother smokes −161.132] [0.980]* [2.31 0.34 −3.42 # Observations 313. Notes: See notes to Table 2.027]+ Mother fixed effects No No Yes 1st trimester pollution 0.130] [3. 16+).22 −0.724 Observations 20.304] [0.501 20.050] [4.344]** [0.0265]** [0.435 20.433 C.574] [2.413 −0.129 [32. * at the 5% level.23 −3.46 −0. these estimates suggest that infants [1] [2] [3] at higher risk of poor outcomes for other biological reasons face A.589 312. 25–34.239 285.69 −3.678] [3. 3.589 [1] CO [2] Ozone [3] PM10 B.58] [17. # Observations 312.447] [6.450] 1st trimester pollution −1.624] [3.239 285.496] 3rd trimester pollution −39.5 −38.401] [0.117 262.714] 1st trimester pollution −0.453] [0. Along cients from models including CO as pollutant in Table 2).33 −15.773 Monitor * quarter fixed effects No Yes Yes [2.482]** [0.69 [5.384]* [0.635]* [1. / Journal of Health Economics 28 (2009) 688–703 Table 2 Effects of air pollution on health at birth— All mothers < 10 km from a Monito.325] [8.12 0.12 [2.05 0.530 305. other race).014 −5.17 −1.485 269.04 −7.009]+ [1.509] 2nd trimester pollution 10. 4+).194]+ 313.11 0. Models of low birth weight (coefficients and standard errors multiplied by 100) 3rd trimester pollution 0.209 The point estimates are substantially larger for very young and very [17. month dummies.504 313. Models of low birth weight (coefficients and standard errors multiplied by 100) # Cigarettes per day −1.87 −2.35 −0.589 312.343]** [0.041 [2. average precipitation and daily minimum and maximum temperature in each trimester before the birth.920] old mothers and for births that had other risk factors.837 1. Table 2).8 −161.008] [3.165 −1.343]* [7.92]* [9.04 0.955] [1.252] [0.11 3. Models of gestation (coefficients and standard errors multiplied by 100) rable to those shown in the main specification (column 3. incompetent cervix.98 −1. and ** at the 1% level. previous preterm infant.196 0.255] [1.65 6. Models of birth weight [0.11 0.14 −0.66 2.227]** [0.19 0. maternal race (African American.117 276.206 C.386] [2. 13–15 years.504 [34.144] [0.530 305.5 −14.18 −3.671 0.846] [4.465] [7.701 268. hydramnios.221]+ [2.227] [1.66 −1.530 305. birth order (2.19 −3.33 2.22]* 1st trimester pollution 0.846 1.24 −6.938]+ in Table 2.31 0.561]+ [8. [1] CO [2] CO [3] CO [4] Ozone [5] Ozone [6] Ozone [7] PM10 [8] PM10 [9] PM10 A. Models of birth weight 3rd trimester pollution −11.32 3.504 313.228]** [0.504 269.07 −1. male child.66 [5.1 −24.94 −13.243 Effects of air pollution on health at birth—all smoking mothers<10 km from a mon- [0.01 4.530 262.05 0.141] [0.35 −1.15] [17. clustered on the census tract level.36 −0.116] [0.0265]** [0.78 [18.14 −3.89 −11.186] 1st trimester pollution −0.31 −3.510] [7. and year dummies.0393]** 2nd trimester pollution 10. Table 3 there do not seem to be stronger negative effects of pollution on Effects of smoking on health at birth—all mothers <10 km from a monitor (coeffi- African-American.504 313.589 268.118] B.41 # Cigarettes per day 0.023]* [1. 2nd trimester pollution 20.769] [2.647 1.317 −15.824] [4.355] [2.708 4.276 1.219]* [1.27 [17. Currie et al.06 4.310] [0.480] [1.70 −1.753]* [3.812] [2.166] [3.494 Risk factors are anemia.117 262.193] [0.273] [2.127]* [1. 3rd trimester pollution −42. heart or lung disease.17 [6.352] [1.06 [2.497 A. + indicates statistical significance at the 10% level.45 −7.20] [14.239 B.247] [0. .171 −0. Models of gestation (coefficients and standard errors multiplied by 100) [30. Models of gestation (coefficients and standard errors multiplied by 100) 3rd trimester pollution −4.83 −0.11 0.464 20.124] [1.57] [9. hypertension (chronic or pregnancy associated).15 −0. hemoglobinopathy.346] [0.504 313.71 0.233]** [1.582] [1.14] [9.465 20.083 C.07 [0.188] [0.144] [0.18 0.530 305.10 −0.485 269.206 286.636 Notes: See notes to Table 2.613] Observations 305.185] 15 Observations 19.10 0.59 −31.08]+ [0. renal disease. Models of low birth weight (coefficients and standard errors multiplied by 100) Mother smokes 4.15 −8.22 −3.90 0.599] 2nd trimester pollution 3.196] [0.137]* [0.691 276.37 −32. median family income in the Census tract in 1989.19 −18.96 −0.620]** itor (mother fixed effects models only).113] [0.251] [0. with the results for smokers. less educated.81 −16.589 312.530 2nd trimester pollution −4.34 [0.492] [3.38 5.429 [0.249]* [1.11 −1.98 −4.91 0.352]** [8.603]+ [3.07 [2.503] [7.77 2. uterine bleeding.43 0.206 286. or “other complications”. previous Mother fixed effects Yes Yes Yes large infant.698 J.311]+ [1.82]+ [15.29 3.89 [6.164] [1.45] Mother smokes −31.782] [2. or low income mothers.118 19.15 [0.245]+ [0.49 −0.981] [3.262 0. dia- Monitor * quarter fixed effects Yes Yes Yes betes.863] [3.00 0.982] [1.542]+ [3.196 0.701 268.150] [0.282]* [0.478] Observations 312.162] [0.701 285.800]** [2.785] [2.448] [7.930 20.234]+ [0.19 −0.521]* [6. eclampsia.339]* [2. and maternal smoking as well as the number of cigarettes per day. These coefficients are from the models in columns (1)–(3) [1.66 −36.081]+ [0. Observations 20.13 −2.28 −2.183] 2nd trimester pollution −0.18] [13.691 Monitor*quarter fixed effects No Yes Yes No Yes Yes No Yes Yes Mother fixed effects No No Yes No No Yes No No Yes Notes: Standard errors in brackets.15 −2.95 4.15 However.10 0. 35+) education (high school.319]** [1.950] # Observations 305.48 0.691 276.08 0.15 −0. marital status. Regressions also include indicators for missing values of the control variables.34 −0.265]** Table 4 # Cigarettes per day −5.78 −7.70 1.124] [0.191] [1.375]** [6.25] [17.11 −4.914]+ [4.22 −19.485 286.482]** [0.793]+ [2.589 312.797]** [4.57 −3.102]+ 1st trimester pollution −14. herpes. multiple birth.667 3rd trimester pollution 4.518]+ [2. rh-sensitivity. Hispanic. Models of birth weight higher risks from pollution. All regressions include indicators for maternal age (19–24.41 3.

000 births reported in Currie and Neidell (2005).7 −5. we control for the zip code of residence.530 Notes: The columns show specifications that allow the effect of pollution to vary by characteristics of the mother.601 7.88 −2.553 −23. suggesting the effects of CO death in the family results in very small sample sizes.129] [5.589 312.567] [0. This result also bolsters offer two investigations of this claim.378 [5. Table 7 shows that this is indeed the case: we socioeconomic factors but has not taken out a greater sensitivity to do not find significant effects on health at birth (or.504 313.530 305. In Table 7 we ical risks but not by non-biological risks. In the last three columns of Table 7.476 [8.92 −11. (4).530 305. find any statistically significant impacts of ozone and PM10 on ozone. The main effects (not shown) are comparable to the main effects in the corresponding specification in Table 2.703] [5.609]+ [0.138] Observations 305.6 averted deaths per 100. . Since we control that are statistically insignificant.52 −37.16 This estimate is remarkably similar to the 16. itors (within 5 kilometers).192]* [4.55] 2nd trimester pollution 6.690] [0.570] [0.18] [10. In this sure to PM10 or ozone. the estimates for CO are even stronger than those with at least one death.718] [0.863] [6. In these models. Currie et al. we estimate models that include both CO and ever.530 305. Table 6 suggests that high CO exposures in less precisely merged sample we find generally smaller estimates the first 2 weeks of life increase the risk of death.589 312.14 [12. J.82]** [11.103]* [4.18 a 1 ppm decrease in CO. how.322 0.6 [12.5 averted deaths per 100.131 9.081 1. Therefore.701] [0. as we once again find significant effects is about 40% (2334 deaths divided by 5848 births).448 0.37] [13. but the pollution measures are interacted with a dummy variable for the characteristic of the mother. Table 6 shows estimates of the effects of pollution on infant mor. As discussed above. with only the interactions shown.589 312.014 −3.888] [6. Both of these results are consis- for the fact that more deaths occur in the first 2 weeks with our tent with improved measurement from knowing the mother’s exact baseline hazard.99] [13.692 −5.08 −9. We do not.479]* [5.504 313.16] [11.009] [4.589 312.695] 1st trimester pollution −4. the resulting reduction in sample size increases our standard errors substantially.754] [0.88 effect of ozone on gestation. A.73 7. Table 6 suggests that CO matters. making it more difficult to draw a clear inference from this exercise.504 313. Unfortunately.81 −24. Models of birth weight 3rd trimester pollution −20. In Table 8. within that time is hastened by pollution exposure.530 305.28] 1st trimester pollution −9. Consistent with the results tors near the zip code. The models are estimated as in Table 2. [6] Income <30.075 1.563] Observations 313.221] [5. Models of low birth weight (coefficients and standard errors multiplied by 100) 3rd trimester pollution 1.97] [12.89]* [9.47] [10.29 1. studies often do not have an exact address of the mother but only tality from models based on Eq.504 313.79] [12.06 −0.119 −4 −4. and therefore assign pollution to the zip birth weight with a series of indicator variables to isolate the effect code centroid using an inverse distance weighted average of moni- of pollution after the birth on health.616] [0.651] 2nd trimester pollution 6. then being farther from a monitor should mother fixed effects has taken out the main effect of confounding yield weaker results.429 −12.447 12.05 −15. Hence.89] [12.118 [15. this estimates reflects the extent to which death address. on pollution that is linked to biological factors.107 0.186 −10. [1] <age 19 [2] ≥age 35 [3] Risk factors [4] Black [5] <12 years ed.504 313.545]* [0. pollution to the mother assuming we only know her zip code.581 −1.345 1.384 [6.638 20. This last trimester rather than the second trimester which seems to calculation suggests 17.736] [11.03 0.108 −0.589 B. we need to account for pollutant drives our results. we believe that a major contribution of our 17 We have also estimated models using mothers who are closer to pollution mon- study is that we can improve the accuracy of our pollution mea. so the base risk of death in this subsample shown in Table 2.438 −5.423 1.144 −9. rather than expo.129 −11. 18 We also estimated our models including an interaction with CO and an indicator 16 We do not show separate estimates of the effect of pollution on deaths among for years after 1995 (midway through our sample) to assess if the effects change infants of smokers because restricting the sample to smokers who had at least one over time. but the interaction term was insignificant.504 C.716] [0. sures because we have the mother’s exact addresses. often therefore vary together.981] [5.245 1.698] 1st trimester pollution 1. infant mortality) for mothers 10–20 km from a monitor.697 0.15] [11.865] [5. Although the sample size is some- the fact that we estimated the impact on the sample of mothers what reduced.920] [0.327 −0.27 −1.18 0.640]** [0.573] 2nd trimester pollution −1.690] [5.17 Similarly.545] [0.785 [0.20] [10. We also find a negative multiply our estimate by the ratio of the overall sample IMR of 6.000 births from matter.767 1. For example.729 [0.575 [0.589 312.000 for the preg. though now it is exposure in the per 1000 births to the subsample IMR of 399 per 1000 births. the effects of pollution appear to be amplified by biolog. If being closer to a monitor the case that our identification strategy is working: including the improves measurement.815 −0. not shown. we of CO on all three infant outcomes.83 −10. are constant over the period.653]+ [0.530 305. / Journal of Health Economics 28 (2009) 688–703 699 Table 5 Effects of CO on health at birth—mothers from vulnerable groups <10 km from a monitor models with mother fixed effects. in the second column the regression include each trimester CO measures interacted with whether the mother is under age 19.15]* [10. Since the sources of these pollutants are similar and mortality.256] [5.47] Observations 312. it is important to isolate which To gauge the magnitude of this estimate.562] [5.343 4.694]* [5.74 −6. Models of gestation (coefficients and standard errors multiplied by 100) 3rd trimester pollution −11.411 0.75] [11.554]* [0. we assign discussed above.

236] 2nd trimester pollution 0. Observations 192.142 0. / Journal of Health Economics 28 (2009) 688–703 Table 6 Table 8 Effects of air pollution after birth on the probability of infant death all mothers Effects of air pollution on health at birth—all mothers <10 km from a monitor models <10 km from a monitor (coefficients and standard errors multiplied by 10.16 −2.404 8.0]+ [525.695] [4.1 16.152 −0.594]** [0. Models of low birth weight (coefficients and standard errors multiplied by 100) 3rd trimester pollution 0.81]** [26. with our richest model identified [1.385] [2.0659 −2.6 −664.414] [10.459 [0.032] [738.20] [25.303 0.0535 0.274 −10.059] [2.620]* [0.6] 1 (Week after birth ≥4) 75.97 [36.358 275.700 J.8 [4.6 278.207] [0.811] [4.228] [0.000). Currie et al.478 −6.646 −0.149 [0.266 −0.605 137.384 0.16 −3.8]** 1 (Week after birth ≥2) −178. All models include mother fixed effects.88 9.275 −0.5] [523.041] [3.71 1.5 [55.01 ppm) −5.28 −5.241] [2.786]** near particular monitors.842 [4. [1] CO [2] Ozone [3] PM10 [1] Birth [2] Low Birth [3] Gestation weight weight Mean pollutant weeks 0–2 101.257] [2.651] [9.0585 4.939 −3.087 [7.044 Mean pollutant weeks >6 −8.551 1.4 −620 −1041.261 3. Coefficients and standard errors are multiplied by 100 in columns [739.736 1.7 1671.01 ppm) 0.195] [1.317 0.964 7. This paper examines the effects of air pollution on [4.618 0.429]* [4. historically low levels of air pollution are likely to 1 (Week after birth ≥12) −1.079] [8.662] [3. Standard errors are clustered on the census tract level.156] [4.11]** [18.117 −1.668 137. .252 1.84 In order to begin to evaluate the costs and benefits of tighter [23.69]** [57.02]+ [4.241] [0.917 for many potential confounders.272 5.207 [9.803 ings are germane to the current contentious debate over proposals Notes: See notes to Table 2. Mothers >10 km and <20 km from a monitor Assigning pollution using zip code [1] CO [2] Ozone [3] PM10 [4] CO [5] Ozone [6] PM10 A.427] [0.416 Mean pollutant weeks 2–4 −21. For example.937] [4.681] [7.395] [4.69]** [62.993] [2.122 [9.273] [0.388 [0.737 3.242] [0.20]** [111.254] [0.952 133.47 −2.365 0.24 79.028 263.613 −14.5 [97.104] [740.0303 −0.192] [2.775 0. Since Number of deaths 2.044]* [15.184] [4.3] 2nd trimester ozone (in 0. the state of California’s most recent proposal to increase emis- Table 7 Effects of air pollution on health at birth—alternative ways to assign pollution.117 276.546] 2nd trimester pollution −11.814] [8.177 8.1 Notes: See Table 2. Week after birth −1713.413 [4.195] [4.849] [2.07 affect health.392 131.681] Observations 248.765 0.957 −4.230 270.406 0. Our models control 1 (Week after birth ≥20) 1.194] [10.51 5.6]* [370.93] 2nd trimester CO (in ppm) 7.366 Mean pollutant weeks 4–6 12.1] [381.314] from current.445]+ [3.6 −908.353 −8.536] [5.504 269.769] Observations 243.6 −3.829] [5.172] [0.347 [4.184 163.087 [9.34] [29. Models of gestation (coefficients and standard errors multiplied by 100) 3rd trimester pollution −0.0]* [369.591 Birth weight 1500–2500 g −1515.962 1.449 −7.163 271. control for both CO and O3.6 [4.79 −5.6 [4. Discussion and conclusions [83.55]** pollution regulation.464 −7.11 −5.696 −3.203] [0.25 2.298 [0.232] [2.0]+ 1st trimester ozone (in 0.23 0.85 0.61 4.43 0.097 −6.379] [1.77 1.917] [5.530 262.888] using variation in pollution between births among mothers located 1 (Week after birth ≥32) 1.2]* [371.263 −0.156 −0.492] [0.434 0.174 2.3 −483.269] [0.8 1643.624] [5.193 267.242] [1. to further tighten automobile emissions standards.8]** [121.206 C. Models of birth weight 3rd trimester pollution −1.082 [10.419] 1st trimester CO (in ppm) −5.47] [9.0]* 2 and 3.215] [0.784 5.239 B. these find- Number of mothers 2.507] [4.1]* Observations 274.167 0.55 41.447 −0.941] 1st trimester pollution −17.410] [5.849 Birth weight 2500–3500 g −1637.463]+ [0.589 268.359] [1.223 −0.691 Notes: See notes to Table 2.870 most CO emissions come from transportation sources.942] 1st trimester pollution −1.846]+ [9.814]+ [3.078 4.474] [0.819] [3.543]+ [5.556 negative effects on infant health both before and after birth.973]* [0.121 0.892] [2.443] [0.271 −0.246] [0.762 −1.208] 1st trimester pollution 0.4 −1066.58 21.463] [0.241 0.157] [737.493]+ [1. it is necessary to understand how changes 1 (Week after birth ≥8) −0.64]** 1 (Week after birth ≥1) 1814. All models include mother fixed effects.9 8. All models include mother fixed effects.9]+ [525.624] [0.561 −1.837 Our strongest and most consistent set of results show that CO has Number of births 5.506] [0.809] 3rd trimester ozone (in 0.210]+ Observations 249.346 0.034] 2nd trimester pollution 0.056 −9.77] [5.818 Birth weight ≥3500 g −1685.115 Birth weight <1500 g −804.7 −1710 −1776.748 313.723 305.334 2.01 ppm) −4.467] [4.42 [8.564 [8.58] 3rd trimester CO (in ppm) −20.69] [109.387] [2.293 2.246 0.701 285.038 1.028 1.411] [1.089] infant health using recent data from New Jersey.3 −122.394 −4.537] [4.146 0.8] [533.478] [0.991 −0.485 286.57]* [97. The models in columns [4] to [6] assign pollution to the child assuming we only knew the mother’s zip code of residence and computing the inverse distance weighted average of monitor values from the zip code centroid.848 5.123 312.109] [1.464] [2.

As such. Unlike small-scale infant death of about 2.000 births. It earnings. We also find that infants of smokers are at much greater as used by the EPA. when it occurred during pregnancy. (Maynard. about the type of exposures that are most likely to be harmful (in mated impact of third-trimester CO on birth weight (−16. at itors. Currie et al. then mean levels of birth weight and gestation will be in annual average 8-h CO concentrations from 1989 to 2003 per increased. or cause state in 2003 to get the change in earnings per birth cohort per state women who might have had low birth weight babies not to become from a 1 ppm change in CO.2 billion in annual risk of negative effects from CO exposure. If time workers per state in 200320 and the total number of births per high levels of pollution cause vulnerable fetuses to be lost. epidemiological studies that use personal air quality monitors To value the impact of recent declines in CO throughout the U. our study is based on the population of live births. This gives us a total of 449 deaths averted in 2003 by the birth weight and gestation suggest that this reduction had an effect reduction in CO from 1989 levels. Finally. When that argument was dis. While we recognize the strong assumptions behind these cal- particularly among smokers. we use a crude proxy for individual exposures. Should the state prevail. we compute the present discounted value of the may still provide a useful benchmark for assessing the benefits of annual earnings increase assuming a 6% discount rate and 30 years further reductions in air pollution in terms of infant health. We further find that a one unit decrease in substantial benefits from the improvements in CO over time. The state is currently suing likely larger than is indicated by the mean decline over the mon- the federal government over the issue. 20 21 Available from the Bureau of Labor Statistics at http://www. While it is not possible to properly assess this. savings. though these estimates are less robust culations. there are several reasons why our estimates may understate 100. We compute the benefits from roughly equivalent to getting a women smoking 10 cigarettes a day these avoided deaths using a value of statistical life of $4. Barringer.S. we do note that the marginal impact of CO on infant mortality we estimate here is virtually identical to the marginal impact of 17 deaths per 100. We also find some evi. Furthermore the decline in actual exposure was meal approach offered by the state. which gives us an estimated increase in Agency. the agency then denied bound. they cohort.. In order to value the improvements in infant mortality. we perform the following illustrative calculations. we regard these estimates as lower state to obtain the increase in annual earnings for the 2003 birth bounds on the benefits of pollution control to infants. 2008). Our estimates of the effects of CO on avoided. 2007.21 dence of significant effects of PM10 and ozone on health at birth. which represents a reduction in the probability of the health impact from pollution exposure. This is clearly a lower missed by the Supreme Court in April 2007.5%.gov/cew/. J. For all these reasons. And since the literature does not give much guidance in birth weight from a unit change in pollution by dividing the esti. We multiply or intensity of exposure) it is possible that more precise measures this by the estimated elasticity between birth weight and earnings taken at key points in the pregnancy would uncover larger effects. grow in the future. which yields an estimated $2. Full details of these calculations are available from the authors’ upon request. / Journal of Health Economics 28 (2009) 688–703 701 sions standards has been blocked by the Environmental Protection of labor force participation. people tend to live in the more least 16 other states are set to implement California’s regulations heavily polluted areas that experienced the largest declines. the mean birth weight in our sibling sample (3236).6 lives saved per 100. since at least in New Jersey. since the assumed discount rate of 6% is relatively high and California’s request for the waiver necessary to implement its law.bls. of 0. strapped to persons. 19 For these calculations we assume a homogeneous relationship between pollu- tion and birth weight or infant mortality. More- mean CO levels in the first 2 weeks of life saves roughly 18 lives per over. Some change in CO by the number of births per state and the decreases areas in our study saw a reduction in mean CO levels from 4 ppm to in CO levels per state to obtain the nationwide number of deaths 1 ppm over our sample period. .000 births found in Currie and Neidell (2005).1 from Black et al.65) by terms of length of exposure. We then multiply this by the average earnings of all full is possible that pollution causes fetal losses or it impairs fertility. We then multiply this by the change pregnant.19 To value the Our noisier measures of exposure may lead us to falsely accept a null improvements in birth weight.8 million to quit. the magnitude of these benefits suggests potentially than our CO estimates. The Agency first argued that it had no authority to regulate nationwide earnings of $720 million for the 2003 birth cohort due the greenhouse gases in auto exhaust. we compute the percentage change hypothesis. we mul- It is noteworthy that we find negative effects of exposure to tiply our estimate of 17. we ignore the fact that mean earnings for this cohort will certainly claiming that uniform federal standards were superior to the piece. (2007) to obtain the percentage change in Furthermore. to the fact that CO had fallen from 1989 levels.000 births for a 1 ppm CO even at the low levels of ambient CO currently observed.

34 30.40]+ Mother married 86.321 3.691 [5.360] [0.003]* [1.338] [0.099]** [7.576 7.702 J.0704 0.027]** [3.31 354.8]* [416.746]** [9.855 −0.5] [371.6 −199.528]** [7.770]** [7.677 −0.19 35.8 −49.608] [0.01 68.979 [0.2 −193.0736 0.1]+ [415.310] [0.444]+ [0.74]** [17.0284 0.26 −0.62]** [92.892]** [4.0419 0.3 56.300]** [4.36]** [89.79]* [8.272]** Mother Hispanic −43.54 50.904 0.6]+ [433.361 −0.69 44.8 [2.336]** [2.379 5.9 −196.651]** [5.12]+ [58.923]** [14.8 −1029.418] Multiple birth −1029.66 112.347] [0.400] [0.42] [51.521] [0.0329 −0.11 [8.293] [0.279] [0.79 −24.863] [3.204 0.348]** [6.738 5.36 −56.846] [4.315]* [0.245 −0.56 49.483]** [0.1 87.702] Mother age missing −197.750] [0.31 49.291]** [7.606]** Male 114 113.01 −43.53 91.67 90.9 9.9 [4.778]** [14.523]** [9.9 −1011.301]** [4.503]** [2.753]* [3.29]** [15.429]** [5.89 −156 −156.812] [2.000 3.554]** [12.337] [0.5 −940.343] [0.02 −0.8 −179 −90.018 2.94 49.89 106.546] [0.555]** [10.718 −0.001]** [8.86]** [23.771] [4.67]** [7.518]+ [2.735]** [6.459]** [0.975 −4.49]** [29.94 −13.34 −3.08 33.3 [6.27 72.94 90.844 [1.4 −39.338] [0.0115 [0.799 0.308] [0.794] [1.347] [0.333] [0.183 [0.64 57.5 −115.695 −1.2 103.06 −6.69 [3.73 92.93 −61.360] [0.447] [6.57]** Birth order 2 87.303] [0.501]** [9.298]** [9.51 −42.108]** [4.564 0.050] [4.39 −39.450] 1st trimester pollution −1.36]** High School 27.842]** [7.414 0.778 −0.273]** [7.159]** [2.558]** [4.9 −1009.85]** [85.554 −0.50]** [20.62 63 36.7 −41.35 8.471]** [8.343]* [7.269 −0.7 −161.41 64.5 57.151 −0.36 [4. [1] CO [2] CO [3] CO [4] Ozone [5] Ozone [6] Ozone [7] PM10 [8] PM10 [9] PM10 Models of birth weight 3rd trimester pollution −11.20] [77.6 −1004.27 −0.08]** [8.636]** [7.09]** [92.147 −0.63 61.78]** [30.92 51.704]** [5. Effects of air pollution on health at birth .265]** [6.1] [13.503]** [3.421 −0.181 −0.5 −712.482]** [0.57 29.16 [6.14] [38.3]** [502.840]** [6.386] [2.208 −0.480]** [7.4 61. Currie et al.32 −5.17 62.5 −197.509] 2nd trimester pollution 10.776]** Birth order 3 99.016]** [8.25 [0.0491 [0.9 114.315] [0.58 −57.786]** [9.7 −1031.428 −0.226 0.348] [0.589]** [6.283]** [7.48]** [7.25]** [22.423 1.1 103.928 50.6 −175.445]** [6.0783 −0.564] Mean daily max temp first trimester 2.39 −1.58 4.04 −11 85.displaying coefficients on all covariates.124 0.24 −6.3 346.301 −0.636]** [14.387] [0.88 31.139 −3.147 −0.62] Birth order missing 36.09]** [7.3 62.5 32.42 −160.486] [0.03 −5.29 52.36 [4.399]** [8.785] [2.8 37.430] Precipitation first trimester 0.059] [5.8 −204.71 −31.195]** [6.547]** [3.5 115 114.45]** [55.914]+ [4.175]** [9.385]** [6.22] [42.36 103.291]** [10.644] [1.394] Mean daily max temp third trimester 1.6 −143 3.44 −728.0536 −0.574] [2.331]* Birth order 4 or higher 71.312] [0.419]** [8.88 [55.819] [3.325] [8.5 21.11 −293 −290 −171.31 −108 −115.453 −7.331]** [2.5 −38.882]** [9.394] Mean daily min temp second trimester −0.492] [3.296]** [2.559]+ [0.26 0.682]** [3.989]** [6.37 25.238]** [8.412]** [7.349]** [3.049 0.99] [22.315]** [0.301] [0.326] [0.6 [85.796]** [5.23 83.655]** [7.5 −151.81]+ [43.436] [4.984]** [13.81 −16.186 −0.5 −46.243 −5.14]** [91.926]** [4.953]** [4.429] [0.373 0.2 −1028.504]** [0.120]** [7.79 −57.009 4.91 −47.01 63.883] Fraction of people poor in 1989 −197.448 −0.0863 0.242 −0.476]** [0.059 −181.6 −107.65 6.92 57.824] [4.56 45.002]** [7.5 [7.51] [20.332 0.684]** Med fam income 1989 in $10.057 −0.05 −2.975]** [7.572] [0.502]** [0.284] [0.013 [4.6 −1.463 1.4 21.88 15.236]* [6.7 −1029.199 −0.653 −0.88 −3.158]** [9.8 122.841]** [7.177]** [4.372] Mean daily min temp first trimester −0.92 31.293]** [2.53 −113.542]+ [3.10 [21.4 −6.315] [0.328]** Number of cigarettes per day −5.333 2.335] [0.503] [7.51 51.353] [0.89 40.478] Mother age 19–24 40.044 54.308]** [5.947]** [6.466] 16 or more years of education 57.6 −204.101 [0.19 0.1] [506.40]* [86.803]** [6.19 −3.33 8.834]** [2.2 −926.566 −3.014 0.252]** [5.98]+ Male missing −713.867]** [10.312 −3.95 −19.1 −144.2 35.810]** [6.375]** [6.844] [2.845 −5.00816 0.54] [4.6 −1030.561] [0.308] [0.355] [2.2 111.4 −44.172 0.36 63.0421 −0.8 −1681.614]** [0.247]** [10.1 −10.517] [0.328 −0.25 7.980]* [2.11 6.039 −7.519]** [5.393]** [4.322]+ [0.482]** [0.113]** [4.994 1.78]** [51.73 [6.959]** [5.54 55.118 −0.841]** [2.643] [0.459] [0.364] [0.423] [0.63 41.087]** [7.028]** [4.22]** [7.639] [1.668 3.189]** [4.71 [5.7 −175 −151.11 −51.239 −0.8 120.424]** [0.504]** [0.8 120.8 −239.647] [0.788]** [8.139 [3.84]** [30.2 −263 −174.510] [7.738 23.365 1.501 [0.345] [0.662 −1.4 −1.54] [15.865 −2.27 86.769]** Mother is smoking −161.7 −164.503]** [0.852 −0.574] [0.558] [0.954]** [5.566 −5.330] [0.517]** [0.413] Precipitation second trimester −0.53]** [77.64 56.01] [8.4 99.960]** [6.352] [0.79 −36.6 32.008] [3.132 −0.392] [0.642]** [3.4 −43.440] [0.8 – [377.428]** [9.0606 −0.592 −2.1 114.7]* [507.808]** [4.06 87.929 56.555 −0.04]+ [83.143]** [5.368] [0.7]** [671.063 −195.981] [3.641 49.3]* [420.52]* [58.237 0.287] [0.105 1.74 0.40 −173.978 −1.562]** [8.86]+ [76.97 100.392] [0.377 [0.725] [0.28]** [37.517] 13–15 years education 52.6 −97.349]** [8.315] [0.167]** [4.511 [0.134 −0.399]+ [0.620]** [0.0112 [0.341]** [5.51]** [13.264 0.523]** [4.4 20.62 40.99 23.47 27.8 −51.176]** [4.166] [3.352]** [8.99 36.6 −1324.675]* [1.999]** [3. / Journal of Health Economics 28 (2009) 688–703 Appendix A.802]** Multiple birth missing −262.482]* [0.335] [0.16] [14.32] Precipitation third trimester 0.67] [38.292] [0.166]** [2.576 [4.727] [3.913]** [6.08] Mother married missing −240.585 1.290]** [5.708] [0.329]** Mother age 25–34 61.14 69.702]** [7.06 100.254 −0.13 −2.630]** [6.014 −5.4] Mother is smoking missing −102.36]** [43.87]** Education variable missing −35.376] [0.210]** [6.47 [6.154]** [4.333]** [4.98 [7.056 −0.496 25.23] [37.314] [0.089 [38.43 −1.2 −749.795]** [3.8 −161.8 [419.088]** [10.267]** Mother age 35 or higher 62.034]** .200] [4.271]* [0.657 [5.14 32.933]** [7.73 40.3 −148.968]** [7.258]** [6.31 0.219 −0.6 −211.239]** [8.80] [38.8 −1676.521]* [6.4 −457.577] [0.33]* [76.284]** [7.368 0.369] [0.57 [7.437 −0.51 40.906 0.708]** [11.393 1.629]** [9.22 85.015 −0.174 [6.669] Mean daily max temp second trimester 0.503]** Mother African American −198.358 −0.543 3.64 −0.00 0.434] Mean daily min temp third trimester −0.

.C. 2001.D. Currie et al. 1350–1366. EPA Office Salam. Y. 1989. 156–164.. Arden. Xu.. David Lee. Burnett. Chay. 4–53.. 2007. N.16]** [106. Indoor Air quality Handbook. Julia Kelsall. Aryl Spengler. tionship between particulate matter air pollution and birthweight. Greenstone. P.D.J. David. David.. Times.3]** [85. Paediatric 391–396. Margolis. / Journal of Health Economics 28 (2009) 688–703 703 Appendix A (Continued ) [1] CO [2] CO [3] CO [4] Ozone [5] Ozone [6] Ozone [7] PM10 [8] PM10 [9] PM10 Other race or race missing −230. Lencar... Air pollution Preterm Birth in the Environment and Pregnancy Outcomes Study at the Univer- and birth weight in northern Nevada 1991–1999.. weight among children born in southern California Between 1989 and 1993. L. Mendola. R.5 3191. lines.. Discrete-time methods for the analysis of event histories. C. 2008. Schoendorf. United States. 1773–1778. Schoendorf. C. The relationship between selected restriction. NBER working paper 10053. P.K. Logan. Ritz.. R. K. Marc.. Demers. market? The effect of birth weight on adult outcomes. Rogers.J..J. Criteria pollutants. sity of California. Yue. Wang. Basu.. Micheline..77]** [67.html...4 −226. Air pollution and infant health: what can we Association of very low birth weight with exposures to environmental sulfur learn from California’s recent experience? Quarterly Journal of Economics 120 dioxide and total suspended particulates. Archives of Environmental Health 53. T. Michael. December 19. Healthy. J. 1). tion and a steel mill. and partic- consumer/03-co.. 1638–1644. From the cradle to the labor 623–628. adverse birth outcomes in Los Angeles County.. Herring.. 1997. Cowen. 2007. 2000.. 2003b.J.. Matthew.. D. WHO Regional Office for Europe. 493–502. Samet. J. A. 2005. 110–115..P.. T. and human capital development. M. Archives of Environmental Health XLVII. 2007. 1031. Mendola. Frery. Health Perspectives CIX.. 2002. Omaye. Woodruff.. K. Epidemiology XI. M. Schwartz.J. R. Jonathan. T. Neas. R.G.J. Parker. Pope. J. Gerald weather confound or modify the association of particulate air pollution with Teague. Greenstone. Environmental Research LVIII... B.W. month and monitor dummies). Environmental Health 5.D. The cost of low birth weight. S. 2008. 2003a. The New York Almond. P.epa. M. Yuanli.. Steer. Dugandzic. Grillo. www. Li. 61–98. J.. New York.9 3363. ation between low level exposures to ambient air pollution and term low birth 2005.. Environmental Allison. Addy.. P.epa. 2003. Woodruff. R. Koehoorn. 141–157. in childhood. 1982.. Arden. L. 1999.X.J. P. Lancet asthma.D. Hoggatt. Millstein.. Wyzga. Lipfert. Los Angeles. 3.. mortality in the United States 1999–2002. A time-series analysis of any short-term Wilhelm. D. K. V.T.F. B. and the Clean Air Act 1993. T. October. W. Environmental Health Perspectives 113 (5).M. 185–194. Daily mortality and PM10 Economics 122.D. 502–511. L. C. Association between tion L.K.2 [380. gaseous ambient air pollutants and adverse pregnancy outcomes in Vancouver. B. Emma..1 −232. American Journal of Epidemiology 145 (4). Scott Zeger. L.. C. but displays all covariates (except for year. S128–S228.4]** [372. Linda. Infant Mortality. 2005. K.... M. 214–227.T. Laurence Kalstein..L. Ritz. Quarterly Journal of Pope.7 [6. H.. Blot. Sagiv. J. T. http://www. The associ.. Relation between ambient air pollution and low birth weight in the northeastern United States. Maynard. Perspectives. Goodrich. K. A time-series analysis of air pollution and preterm birth in Pennsylvania weight: a retrospective cohort study.. Air pollution and birth Times. 409–439. of 1970... Kjell Salvanes.J. poor health target population? Pediatrics 118 (1).147]** [6. K.C.C. 1995.H. Synergistic effects of air pollution and personal smoking on ology 20 (6). Preterm birth after the Utah Valley Basu. Adolfo. Zhang. Parker. S. Woodruff. 1997. Air pollution and postneonatal infant Canada. Environmental Health UK. Maternal smoking during pregnancy and child birth weight. Girard. Grant. Socio. 2000. carbon monoxide. 2006.2 3438. Fei Yu. Yang.. S. Estimating separately personal prehensive analysis of US data for 1990. Thompson... Relationships between DP0828. Parker. Felicity. S.. Elementary school absences and PM10 pol- study of traffic-related air pollution impacts on birth outcomes. J. M. Guadalupe Chapa. C. Journal of Exposure Analysis and Environmental Epidemiology 3 Tominey. C. Godin. Smith-Doiron.D. 2006. M. Moreau.. 166. 2006. C. 2005.. Air Pollution and infant death in southern Cnattingius.. 176–178. 2001.. Daniel. The New York Parker. Maternal age modifies the effect of maternal smoking on California 1989–2000.. Paediatric and Perinatal Epidemi. 2003. Sven. 1998.6 2626 2825.. 602–613. Currie. 1993.. Inhalation Toxicology 14 (2). WHO Air Quality Guide- 336–338. Epidemiology 19 (6). Ritz. S. Jennison. of Mobile Sources Fact Sheet OMS-3. EPA denies California’s emissions waiver... Barringer. Michael. Huel. Lurmann. Mage. 351–356. http:// Perspectives 113 (11). 211–216. 17–25.. California USA. Filippi.. Impact of changes in transportation and commuting behaviors mortality? Environmental Research Section A LXXVII. American Journal of Public Health LXXIX. Health Perspectives 116 (5). 1953. Woodruff.gov/oar/oaqps/greenbk/o3co.343]** [7. risk assessment: why and how? Journal of the Air Waste Management Associa- Liu. Schoendorf.L. 2005.. A. B. 2000. J. W. F..T. 187–189. A.C..J. Y...D. J. Air pollution and very low birth weight infants: a Currie. Bell. 1992.E. Air Quality. Poole. K. 319. Paul. 2006. G. K. Comparing steel mill closure a natural experiment. weight among term infants in California. Douglas. Respiratory disease associated with community air pollu- exposure to air pollution? Epidemiology 18 (5).F. L.. Chay. 1167–1183. S. Is risk of low birth weight affected by maternal Pope. 2nd ed. J. Samet. Jonathan. Effect of air pollution on Quarterly Journal of Economics CXVIII. Brauer. M. J. 2004. 2007.. Journal of Exposure Analysis and Environmental Epidemiology 14 (5).. Karr. 454–461....5 2947. Janet.. Jing Xu. 820–823. Joel. Parker. C. evidence from geographic variation in pollution shocks induced by a recession.E. Environmental Health Perspectives Liu. 9–19. 1045-52. McCarthy. Environmental Health Environmental Protection Agency.. pollution in Utah Valley. 204–219. Lancet I 1952. Woodruff... W.56]** [107. air pollution and preterm birth in California. Ghosh. Wilhelm. D. J. Stieb. Kenneth Powell. intrauterine growth retardation but not on late fetal death and placental aburp. Rogers.1]** Observations 312589 312589 312589 268701 268701 268701 285239 285239 285239 Monitor* quarter fixed effects no yes yes no yes yes no yes yes Mother fixed effects no no yes no no yes no no yes Notes: The table corresponds to the same regressions as Table 2 Panel A. Neidell. Association between mater. Glasg. C. Automobiles and carbon monoxide. Black. Quarterly Journal of Economics 121..061]** [7.. LeRoy Graham. American Journal of Epidemiology. References Maisonet..4 −234.9 −228. McKeown. Rose. Y. Environmental Health Perspectives CXI.F. Arden.C. M. Lori Hutwagner.310]** Constant 3174. Fei Yu. F. Wilson. (Suppl. 1997. Y.465]** [7.F. 2008. American Journal of Epidemiology 151 (3). William. C. 116 (1). ulate matter: results from the Children’s Health Study.D. 9. P.. Mortality in London fog incident... 2000. 113. B. 2008. and Perinatal Epidemiology 22 (3).. D.7]** [60. Influences of study design and location on the rela- ifornia. 2000. Mildred.. Zhao. Sandra. World Health Organisation (WHO). Shi. Carbon monoxide. Environmental lution in Utah Valley. Ebisu. Anthony.. Jouni. T. exposure metrics in the relationship between Pm2 5 and birth weight in Cal. Nessmann. Chen. 121–128. Dunlop. Particulate air pollution and acute health effects. Pediatrics 115 (1). .1 3319. Copenhagen. CCCLIV. Local variations in CO and particulate air pollution and effects of meteorological and air pollution factors on preterm births in London. Infant mortality and air pollution: a com. Ambient Air Pollution and Chen. T. Journal of the American Medical Association CCLXXXV... Shiliang. Pediatrics 118 (2).3 −228. Hajat.. M. K. F. Does Friedman.D. F. J.A. T.gov/oms/ Birth outcomes and prenatal exposure to ozone.A. the impact of air pollution on infant mortality: Environmental Health Perspectives CVII. Beate. Saulnier.. 1992... J. 2000. February 20.59]** [72. Paul Devereux. Scott Fruin. delivery.. Dodds. January 3. 680–686. Lee. Journal of Exposure Science and Environmental Epidemiology 17 (5). Michael. 1212–1221.897–905. 602–606. Journal of Economic Literature. John.pdf. Environmental Protection Agency.8 3280. Wilhelm.. Kenneth.. wealthy. Lester. Utah Valley.W. Timothy.L. nal exposure to ambient air pollutants during pregnancy and fetal growth Woodruff. Decoufle. January 1993. Woodruff.. 2007. Ritz. K. J.. M. Environmental Research 106 (2). 2005. Ritz. Environmental Health Perspectives CV.J. 2008. adult pulmonary function. Janet.. Journal of the Air & Waste Management exposure to ambient and nonambient particulate matter for epidemiology and Association 50 (8). et al.. Tamburic. Ransom.L.. 608–612. Beate. S. 1997–2001. J. L. 2008. NY. Darrow. Schoendorf. A cohort Ransom. M. Parker. logical Methodology XII. The effects of ambient carbon monoxide on low birth tion. 2007. and wise? Socioeconomic status. Savitz. Belanger.. California sues EPA over denial of waiver.337]** [7.47]** [75. Krewski. Gilliland. Loomis.. causes of postneonatal infant mortality and particulate air pollution in the 426–432. London School of Economics Center for Economic Performance Working Paper Huynh.J. 2008. (6). preterm birth among children born in southern california between 1989 and Chay. J. X. hydrocarbon hydroxylase activity in human placenta and threatened preterm McGraw Hill Book Co. Richard Burnett.J. during the 1996 Summer Olympic Games in Atlanta on air quality and childhood Seaton.