Content uploaded by Sammy Zahran
Author content
All content in this area was uploaded by Sammy Zahran on Oct 21, 2017
Content may be subject to copyright.
The Effect of Leaded Aviation Gasoline
on Blood Lead in Children
Sammy Zahran, Terrence Iverson, Shawn P. McElmurry, Stephan Weiler
Abstract: Lead is a neurotoxin with developmentally harmful effects in children. In
the United States, over half the current flow of lead into the atmosphere is attribut-
able to lead-formulated aviation gasoline (avgas), used in a large fraction of piston-
engine aircraft. Various public interest firms have petitioned the EPA to find endan-
germent from and regulate lead emitted by piston-engine aircraft, though the EPA
has so far ruled against such petitions. To address an EPA request for more evidence,
we construct a novel data set that links time and spatially referenced blood lead data
from over a million children to 448 nearby airports in Michigan. Across a series of
tests, and adjusting for other known sources of lead exposure, we find that child
blood lead levels (1) increase dose-responsively in proximity to airports, (2) decline
measurably among children sampled in the months after 9/11, (3) increase dose-
responsively in the flow of piston-engine aircraft traffic, (4) increase in the percentage
of prevailing wind days drifting in the direction of a child’s residential location, and
(5) behave intuitively and significantly when considering two-way and three-way in-
teractions of our main treatment variables. To quantify the policy relevance of the
results we provide a conservative estimate of the social damages attributable to avgas
consumption. Damages are at least $10 per gallon, which can be compared to a pump
price of about $6 per gallon.
JEL Codes: I120, I180, J130, Q510, Q530
Keywords: Aviation gasoline, Blood lead levels, Child health, Lead exposure
CHILDREN EXPOSED TO LEAD have diminished life chances. Studies link lead expo-
sure to adverse mental and behavioral outcomes, such as IQ loss, poor academic achieve-
ment, attention-deficit disorders, delinquency, and violence and to irreversible physical
Sammy Zahran is at Colorado State University, Department of Economics, and is Robert Wood
Johnson Health Scholar at Columbia University (Sammy.Zahran@colostate.edu). Terrence Iverson
is at Colorado State University, Department of Economics (Terry.Iverson@colostate.edu). Shawn
P. McElmurry is at Wayne State University, Department of Civil and Environmental Engineering
(S.mcelmurry@wayne.edu). Stephan Weiler is at Colorado State University, Department of Eco-
Received April 7, 2015; Accepted September 29, 2016; Published online April 11, 2017.
JAERE, volume 4, number 2. © 2017 by The Association of Environmental and Resource Economists.
All rights reserved. 2333-5955/2017/0402-0007$10.00 http://dx.doi.org/10.1086/691686
575
This content downloaded from 129.082.187.225 on April 12, 2017 14:16:26 PM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
health problems such as hypertensive disorders, damage to renal and cardiovascular
systems, and tooth decay.
1
While lead has been banned in the United States from
the largest original sources—paint, plumbing, food cans, and automobile gasoline—
deposition associated with lead-formulated aviation gasoline (avgas) from piston en-
gine aircraft (PEA) remains an important source of new emissions. The flow of lead
from PEA constitutes between half and two-thirds of remaining lead emissions in the
United States (EPA 2008). Advocacy groups have petitioned the Environmental Pro-
tection Agency (EPA) to find endangerment from these emissions, but the agency has
so far declined, holding that additional studies are needed “to differentiate aircraft lead
emissions from other sources of ambient air lead”(EPA 2010a, 2).
While some studies have linked avgas use to elevated atmospheric lead levels in the
vicinity of airports (Piazza 1999; Tetra Tech 2007; Callahan 2010; EPA 2010b; Carr
et al. 2011), to date only one study has linked airport proximity to blood lead levels
(BLL) in children. Miranda et al. (2011) find a significant correlation between child
BLL and proximity to airport facilities in six counties in North Carolina. While this
spatial correlation is highly suggestive, to more conclusively link lead in avgas to child
BLL, we need to disentangle the flow of lead due to aviation-related sources from other
exposure pathways that potentially increase in airport proximity.
A common exposure pathway for children in the United States is dust associated
with deteriorating or haphazardly removed lead-based paint. Exposure to lead-based
paint is primarily a problem in old houses, particularly in homes built before 1950. In
our study area, the percentage of homes built prior to 1950 is almost twice as high in
neighborhoods proximate to an airport compared to those more distant.
2
Moreover,
due to zoning restrictions, lead-emitting industrial facilities are more common in the
vicinity of airports. Of the 4001census tracts within 2 kilometers of an airport in
Michigan, 41% also have a lead-emitting facility within 2 kilometers. Failure to account
for the spatial coincidence of older homes and point-source polluters could inflate es-
timated health risks from avgas exposure. We address this spatial coincidence problem
by including neighborhood measures of housing stock age and location of industrial point
sources, among other relevant controls.
In addition to accounting for alternative exposure pathways, airport proximity (in
itself) is an incomplete measure of avgas exposure risk. First, airports vary immensely
1. Needleman and Gastonis (1990), Dietrich et al. (2001), Canfield et al. (2003), Nevin
(2006), Miranda et al. (2007), Reyes (2007), Jusko et al. (2008), Zahran et al. (2009), Nigg
et al. (2010), Mielke and Zahran (2012), Reyes (2012).
2. The percentage is 44% for neighborhoods ≤2 kilometers of an airport compared to 23%
for neighborhoods 9–10 kilometers from an airport.
nomics (Stephan.Weiler@colostate.edu). We thank the Michigan Department of Community
Health, Childhood Lead Poisoning Prevention Project for providing the blood lead data used in this
study. We thank the Robert Wood Johnson Foundation Health and Society Scholars program for
its financial support. We thank Mark A. S. Laidlaw for comments on earlier drafts.
576 Journal of the Association of Environmental and Resource Economists June 2017
This content downloaded from 129.082.187.225 on April 12, 2017 14:16:26 PM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
in PEA traffic. In our sample, the average monthly number of PEA operations varies
from 7 (at MTC Selfridge) to 1,099 (at PTK Pontiac). Neglecting the volume of PEA
traffic amounts to assuming that all airports traffic equally in PEA. Second, the fate
and transport of avgas emissions depend on the direction of prevailing winds that vary
in and across airport facilities. Insofar as avgas is an independent source of lead expo-
sure, two children equidistant to the same airport face different risk of elevated blood
lead depending on the child’s residential near angle to the nearest airport. We address
these important sources of omitted variable bias in previous literature by including mea-
sures of prevailing wind direction and PEA traffic.
Despite more extensive controls and improvements in the operationalization of ex-
posure risk, variation in child BLL near airports may pick up more than locational dif-
ferences in lead deposition from PEA aircraft. First, a residential selection bias may
operate if families residing near airports are less prone to undertake defensive ac-
tions, including measures to protect against alternative exposure pathways like lead-
contaminated dust. In Michigan, populations of lower socioeconomic status are more
likely to reside near airports. Compared to more distant neighborhoods (9–10 km),
neighborhoods within 2 km of an airport have significantly higher percentages of house-
holds receiving public assistance (4.35 vs. 8.41, t53:57) and lower levels of educational
attainment among adults (≥high school education, 81.45 vs. 75.98, t5–2:03). In
addition to controlling for these and other measures of neighborhood socioeconomic
status in regression models, our measure of prevailing wind direction provides an exog-
enous source of variation with respect to the problem of residential selection. In Mich-
igan, prevailing wind direction is statistically unrelated to neighborhood socioeconomic
composition at ≤2kmofanairport.
3
That is, disadvantaged populations proximate to
airports are equally likely to be up versus downwind.
Second, an underappreciated source of child lead exposure is lead-concentrated
soils, primarily due to legacy deposition from lead-formulated automobile gasoline. Ur-
ban geochemists have linked child BLL to the accumulation of lead in neighborhood
soils. Contaminated soils enter a child’s body through ingestion (involving hand-to-
mouth behaviors) or inhalation of lead-concentrated soils resuspended during summer
months (Filippelli et al. 2005; Laidlaw et al. 2005, 2012; Zahran et al. 2010, 2011,
2013). Both aircraft traffic and the atmospheric resuspension of contaminated soils
peak in summer and retreat in winter (Laidlaw et al. 2012; Zahran et al. 2013).
4
Fail-
ure to account for this seasonal coincidence of avgas deposition and resuspension of
legacy sources could upwardly bias the evaluation of avgas risk. To address this specific
3. The correlations between downwind risk (measured as the percentage days where wind
drifts in the direction of a neighborhood) and percentage receiving public assistance (p5:62),
median home value (p5:32), and percentage with ≥high school education (p5:29) are in-
distinguishable from chance.
4. In our sample, PEA departures and arrivals are significantly higher (t5–6:43, p<:01)
in the summer (428 per month) than in the winter (286 per month).
Effect of Leaded Aviation Gasoline on Blood Lead in Children Zahran et al. 577
This content downloaded from 129.082.187.225 on April 12, 2017 14:16:26 PM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
problem, our research design exploits two independent sources of exogenous variation
in lead deposition from piston engine aircraft traffic. First, we use an exogenous lead-
deposition shock that resulted from the grounding and restriction of PEA traffic fol-
lowing the tragic events of September 11, 2001. Reflecting the drop in traffic, avgas
sales in Michigan declined over 50% the month after September 11, 2001. Second,
we use monthly data on PEA arrivals and departures by airport to test whether child
BLL is dose-responsive in the volume of PEA traffic. This exploits spatial and tempo-
ral variation in PEA traffic driven by local meteorological conditions that are plausibly
exogenous from other exposure pathways.
The analysis deploys a novel data set that includes BLL records for over a million chil-
dren linked spatially and temporally to 448 fully operational airports across Michigan, and
emissions data for all toxic release inventory facilities that emit lead. In addition, for a sub-
set of airports, we also observe the monthly count of piston-engine aircraft operations.
Across all tests rendered, we find consistent evidence that avgas use is significantly linked
to elevated BLL in children near airports. Child BLL and the odds of eclipsing various
CDC thresholds for concern (1) increase in proximity to airports, (2) decline in the months
after September 11 among children proximate to airports, (3) increase in the flow of
PEA traffic, (4) increase in the percentage of downwind days, and (5) behave intuitively
with respect to two-way and three-way interactions of the main treatment variables.
To quantify the social cost of avgas exposure, we deploy a standard syllogism link-
ing BLL to IQ loss, and IQ loss to future earnings (Schwartz 1994; Grosse et al. 2002;
Gould 2009). We find that reducing PEA traffic in Michigan from the 50th percentile
(407 monthly operations) to the 10th percentile (133 operations) would generate a
social benefit, measured in terms of the net present value of future earnings, of about
$120 million. This translates to a bit over $10 in external social cost per gallon of avgas
sold, which can be compared to a pump price of about $6 per gallon.
5
This estimate
may be regarded as conservative because we consider only deposition near airports on a
subset of the population (children under five), and we only account for the impact of
IQ loss on earnings, one of several known damage channels.
6
1. BACKGROUND: RATIONALE FOR AVGAS USE
AND REGULATORY RESPONSE
Despite recent national interest in lead exposure following the preventable failure of
the water distribution system in Flint, Michigan, lead pollution in the United States
5. Self-service price retrieved for Coleman Young Airport in Detroit, September 1, 2014.
6. In addition to IQ loss, lead exposure can cause growth stunting, seizures, and lasting dam-
age to various body systems. Kemper, Bordley, and Downs (1998) provide comprehensive health
care cost estimates from medical interventions necessary to treat both low and high level expo-
sure to lead. Others have estimated the total direct costs of lead-linked crime, including victim
costs, criminal justice processing, and incarceration, as well as lost earnings to victims and per-
petrators of crime (Gould 2009).
578 Journal of the Association of Environmental and Resource Economists June 2017
This content downloaded from 129.082.187.225 on April 12, 2017 14:16:26 PM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
has evolved for the most part into a legacy problem. BLLs have declined dramatically
in time (Raymond, Wheeler, and Brown 2014), and the most important exposure
pathways for children nationwide involving legacy sources of lead-based paint and con-
taminated soils (Farfel et al. 2003; Zahran et al. 2010). Nevertheless, for the approx-
imately 16 million people—and 3 million children—who live within a kilometer of air-
port facilities that service piston-engine aircraft, the continuing flow of lead into the
environment remains a potentially serious source of exposure risk.
About 160,000 piston-engine aircraft are registered in the United States, consti-
tuting about 70% of the US air fleet. In 2011, these aircraft consumed an estimated
225 million gallons of avgas (Kessler 2013). This consumption implies a flow into the
environment of about a million pounds of lead each year. While small compared to the
amount consumed historically in automobile gasoline, the impact is spatially concentrated
with approximately half of the lead from avgas depositing near airports (EPA 2008).
The primary rationale for the continued use of lead in avgas is aircraft safety. When
engines are run at high power, as they always are in aircraft, there is a risk of pre-
ignition under compression—also known as “knocking.”This damages the engine and
can lead to sudden engine failure. Tetraethyl lead is one of the best-known additives
for avoiding dangerous knocking (Ells 2006). The high intensity at which aircraft en-
gines operate, together with the high stakes of engine failure, explain why tetraethyl lead
is still used as an additive in avgas even though it has been banned from all other trans-
portation fuels. Nevertheless, approximately three-quarters of the existing piston-engine
fleet could transition safely to lead (and ethanol) free automotive gasoline at negligible
additional costs (Kessler 2013). These planes continue to use leaded avgas in large part
because it is the primary fuel available at many US airports.
In 2006 (and again in 2014 and joined by Physicians for Social Responsibility and
Oregon Aviation Watch), the environmental group Friends of the Earth petitioned
the EPA to find endangerment from and regulate lead emitted by piston-engine air-
craft (EPA 2014). While both the EPA and the CDC have recognized that there is no
known safe level of lead exposure (CDC 2012a, 2012b; DHHS 2012), the EPA ruled
against the petition, calling for more studies to substantiate the risks. In 2013, the
Federal Aviation Administration (FAA) announced the formation of the Piston Avi-
ation Fuel Initiative, a joint effort between the FAA and industry partners, with the
expressed goal of finding an unleaded replacement fuel that could be used as a drop-in
substitute for the entire general aviation fleet by 2018 (FAA 2012). The findings be-
low explicitly address the EPAs request for information. Our results lend credence to
the concern of advocacy groups and add impetus to the FAA’s ongoing effort to find a
substitute fuel.
2. MATERIALS AND METHODS
2.1. Response Variables
Blood lead data were obtained by confidentiality agreement from the Michigan De-
partment of Community Health (MDCH). The data set contains blood samples from
Effect of Leaded Aviation Gasoline on Blood Lead in Children Zahran et al. 579
This content downloaded from 129.082.187.225 on April 12, 2017 14:16:26 PM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
1,043,391 fully observed children collected from January 2001 through December
2009 under the Healthy Homes and Lead Poisoning Prevention (HHLPP) program.
HHLPP is funded by the CDC and enlists health providers across the state. The pro-
gram is intended to support lead poisoning prevention and surveillance services for
children in Michigan. Blood lead samples are collected during regular visits to a doctor
with a sampling emphasis on at-risk children residing in older homes or neighbor-
hoods known to have children with elevated blood lead levels. The total number of
blood lead samples collected under the program represents about one-sixteenth of all
children under 72 months of age in Michigan.
Blood lead data are reported in units of micrograms per deciliter of blood (mg/dL).
The MDCH data also contain information on the census tract residential location of
each child, the month and year of sample collection, child age in years (0–5), and child
sex (male 51, female 50). As with previous research (Zahran et al. 2011), we an-
alyze child BLL as a binary variable (≥5mg/dL 51, <5mg/dL 50, and ≥10mg/
dL 51, <10mg/dL 50). Two reasons motivate our decision to use threshold re-
sponse variables instead of a continuous measure of blood lead. First, our thresholds of
≥5 and ≥10 mg/dL correspond to the CDCs present and past reference levels of ele-
vated blood lead. Children with BLLs exceeding these “levels of concern”require case
management. Second, 40.2% of children sampled have BLLs that are at or below test
detection limits. Technically, the precise amount of lead in the bloodstream of a sam-
pled child at or below detection limit is unknown. We can, however, determine with
certainty whether or not a child’s BLL is ≥5or≥10 mg/dL.
7
2.2. Avgas Exposure Variables: Distance, Traffic, and Wind
Point location data on airports in Michigan were gathered from the Geographic Names
Information System (GNIS). A total of 448 airports satisfied our inclusion criterion
of having at least one child (with a BLL reading) residing within 10 km. Distance is
measured in kilometers from the population-weighted centroid of each census tract
where a child resides to the nearest GNIS airport. Distance to a hazardous land use
is a standard proxy for exposure risk (see Rau, Urzua, and Reyes 2015). Following Mi-
randa et al. (2011), we use distance data to test whether child BLLs are dose-responsive
in distance to GNIS airports. For tests involving distance to the nearest GNIS airport,
the sample size of fully observed children is 1,023,672.
We also collected data from the Federal Aviation Administration’s Operations and
Performance (FAAOP) system on the monthly sum of piston-engine aircraft depar-
7. To demonstrate the robustness of our findings (with respect to sign/significance), we also
report statistical results with child BLLs measured continuously but caution about the potential
for error-in-measurement given the noted test detection problem in the data. In models with
BLL treated continuously, BLL is log transformed to address positive skew (S55:70) and se-
vere kurtosis (K591:15). By taking the natural log of BLL, we eliminate skewness (S50:57)
and substantially minimize kurtosis (K52:84).
580 Journal of the Association of Environmental and Resource Economists June 2017
This content downloaded from 129.082.187.225 on April 12, 2017 14:16:26 PM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
tures, arrivals, and aircraft seat count. A total of 27 airports are inventoried in the
FAAOP system in Michigan from 2001 to 2009. The month of blood draw is linked
to the corresponding month of PEA traffic at the nearest FAAOP airport, and we test
whether child BLLs are dose-responsive in the volume of piston-engine aircraft traffic.
Our use of current month without lags is motivated by the following considerations.
Estimates of the half-life for lead in blood vary from as low as 15 days (Manton et al.
2000) to a more typical range of 21–30 days (Rabinowitz 1991; Lidsky and Schneider
2003). We explored several possible operations for lead exposure risk, including cur-
rent month PEA traffic, prior month, and two months prior, as well as 2- and 3-month
rolling averages. Of these, current month is the best predictor of elevated blood lead
risk. Relatedly, figure 2, panel A, of the results section demonstrates a striking concur-
rent correlation between monthly PEA traffic and average BLLs in children. While it
would be tempting to use several such measures in the econometric analysis, doing so
would invite multicollinearity due to significant serial correlation in monthly PEA
traffic across airports. In tests involving the use of distance to the nearest FAAOP air-
port, the sample size of fully observed children decreases to 364,292, corresponding to
the reduction from 448 GNIS airports to 27 FAAOP airports.
In addition to airport proximity and the volume of PEA traffic, child exposure risk
is also influenced by local wind patterns. To account for this, we collect prevailing wind
direction distribution data at each FAAOP airport (from www.windfinder.com). To
illustrate, figure 1 presents a compass plot of prevailing winds at DET, Coleman Young
Municipal Airport in Detroit, Michigan. With near angle information linking a child’s
census tract location to the nearest FAAOP airport, we estimate downwind risk as the
percentage of wind days that drift in the direction of the compass octant of a child’s
residential location. In addition to providing more valid operationalization of exposure
risk, prevailing wind direction provides an exogenous source of variation with respect to
the residential selection.
2.3. Control Variables
The econometric models control for a variety of other sources of lead exposure. Data
from the Toxic Release Inventory (TRI) system identify 578 facilities that emitted lead
or lead compounds in Michigan between 2001 and 2009. We measure the distance
from the population-weighted centroid of each census tract to these lead-emitting fa-
cilities. TRI data allow us to test whether the count of point source polluters within
2 km of a child’s residential neighborhood increases their likelihood of exceeding var-
ious CDC thresholds for concern.
8
8. We calculated various distance buffers (0.5 km, 1 km, 1.5 km, etc.) and determined
through both statistical analysis (in terms of predictive efficacy) and prior research (in terms
of emissions dispersion) that a 2 km buffer was optimal.
Effect of Leaded Aviation Gasoline on Blood Lead in Children Zahran et al. 581
This content downloaded from 129.082.187.225 on April 12, 2017 14:16:26 PM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
To proxy for the risk of lead-based paint exposure, we use census tract housing
data from the US Census Bureau measuring the percentage of housing stock built prior
to 1950. Following Miranda et al. (2011), we also measure the percentage of house-
holds receiving public assistance income, median home prices, and percentage of adult
population with a high school education or greater to estimate the socioeconomic con-
ditions of a child’s neighborhood (i.e., census tract). Studies show that children of low
socioeconomic status are at greater risk of presenting with elevated BLLs (Campanella
and Mielke 2008; Zahran et al. 2009). Finally, we also measure population density
since this correlates strongly with road density, and road density is a reasonably good
proxy for prior period use of leaded automobile gasoline and consequent accumulation
Figure 1. Prevailing wind direction distribution at DET (Coleman Young Municipal Air-
port in Detroit, Michigan).
582 Journal of the Association of Environmental and Resource Economists June 2017
This content downloaded from 129.082.187.225 on April 12, 2017 14:16:26 PM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
of lead in neighborhood roads and soils (Quinn 2013; Zahran et al. 2013). Table 1
reports descriptive statistics for all variables.
2.4. Econometric Models
To analyze whether child BLL is dose-responsive in airport distance, we estimate a
series of random intercept generalized least squares and logistic regression models with
child BLLs measured continuously and dichotomously. The use of time-invariant co-
variates (like airport distance) necessitates the use of random as opposed to fixed ef-
fects regression. For reasons discussed in the previous section, our analytic emphasis is
on threshold response variables instead of a continuous measure of blood lead. We
therefore restrict our presentation of reduced form equations to logit models. All re-
gressions include a tract-specific random intercept (z
j
) to account for unobserved char-
acteristics or conditions at the tract scale (like the accumulation of lead in neighbor-
hood roads and soils). The term Yindicates a BLL surpassing a given threshold for
Table 1. Descriptive Statistics
Variable Mean SD Min Max
Blood lead (mg/dL) 2.98 3.00 1 164
≥5mg/dL .16 .37 0 1
≥10 mg/dL .03 .17 0 1
Distance GNIS airport (km) 4.35 2.13 .11 10.00
Distance FAAOP airport (km) 6.02 2.29 .47 9.99
PEA traffic (monthly) 406.09 267.06 1 1,766
Downwind (%) 12.04 4.53 1.7 31.85
Age 2.25 1.43 0 5
Male .51 .51 0 1
Winter .19 .40 0 1
Spring .23 .42 0 1
Summer .29 .46 0 1
Fall .28 .45 0 1
Housing built <1950 (%) 37.34 24.38 0 100
Population density 4,419.77 3,855.84 2.21 19,392.68
Public assistance (%) 6.65 5.94 0 30.79
≥high school education (%) 77.12 13.43 0 100
Median home price ($10,000) 9.27 5.29 1.43 85.4
Pb facility ≤2 km .88 1.59 0 13
Year 2005.50 2.49 2001 2009
Note. Descriptive statistics on age of child, blood lead outcomes, and season of blood draw are reported
for all fully observed children (N51,023,672) in Michigan from 2001 to 2009. Distance to nearest
FAAOP airport, PEA traffic, and downwind descriptive information pertain to the subset of airports in-
ventoried in the FAAOP system.
Effect of Leaded Aviation Gasoline on Blood Lead in Children Zahran et al. 583
This content downloaded from 129.082.187.225 on April 12, 2017 14:16:26 PM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
concern; Y51 if blood lead is ≥5mg/dL (or ≥10 mg/dL), and Y50 if blood lead
is <5mg/dL (or <10 mg/dL); Yis modeled, for child iin census tract jand month t,
by the following reduced form logistic equation:
Prob Yijt 51jDj,Mi,Ai,Zt,St,Fj,Hj,Pj,Ij
5L½aij 1β1Dj1G1Mi1G2Ai1G3Zt
1G4St1λ1Fj1λ2Hj1λ3Pj1λ4Ij1zj:
(1)
Here, Λ[·] is the cumulative distribution function of the logistic distribution, a
ij
is the
model intercept, corresponding to the likelihood of a reference child in census tract j
eclipsing a CDC-defined threshold, D
j
is the distance (in km) of the population-weighted
centroid of census tract jto the nearest GNIS airport, Mi51 if the child is male, A
i
denotes a series of dummy variables corresponding to child age in years, Z
t
is the year
the blood sample was drawn, measured as a series of year dummy variables with 2001
as our reference year, S
t
is the season a blood sample was drawn, F
j
is an indicator var-
iable that equals 1 if a lead facility operates within 2 km, H
j
is the percentage of hous-
ing stock in a child’s neighborhood built before 1950, P
j
is the population density in
the child’s neighborhood, and I
j
is a vector of neighborhood socioeconomic character-
istics, including the percentage of households in a child’s neighborhood receiving pub-
lic assistance income, percentage of adults ≥a high school education, and median home
prices. In addition to measuring distance continuously, we examine categories of dis-
tance (<1 km, 1–2 km, 2–3 km, 3–4 km, with >4 km constituting our reference cat-
egory) to check for nonlinearities in the relationship between child BLL and airport
distance. This first test is meant to reproduce Miranda et al. (2011), with the addition
of more covariates (accounting for other sources of lead exposure). Insofar as deposi-
tion of lead from piston-engine aircraft traffic is a source of blood lead in children, we
expect the odds of a child eclipsing CDC reference values to decrease in distance from
GNIS airports.
Our next test is designed to separate the flow of avgas from the stock of lead in the
lived environment that circulates seasonally (see Laidlaw et al. 2012; Zahran et al.
2013). Following the tragic events of 9/11, aircraft traffic in the United States was sub-
stantially restricted. The effect of this aircraft traffic restriction is reflected in monthly
aviation gasoline sales and deliveries, which were significantly lower than expected in
September, October, and November of 2001.
9
Insofar as avgas sales proxy for the
9. We estimated the following fixed effects least squares model to observe the deposition
shock effect: ln(Git )5β01β1shock 1G1month 1G2year 1εi, where G
i
is the sale of avia-
tion gasoline by prime suppliers in Michigan (in thousands of gallons) in month i,shock is an
indicator variable 51 if the observation is from September to November in 2001, month is
a suite of monthly dummy variables (with January as our reference month), year is a suite of
584 Journal of the Association of Environmental and Resource Economists June 2017
This content downloaded from 129.082.187.225 on April 12, 2017 14:16:26 PM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
monthly level of lead deposition across GNIS airports, we analytically leverage the ex-
ogenous restriction of PEA traffic as a quasi-experiment in lead deposition. In the air
traffic restriction period following 9/11, avgas consumption drops markedly, while at-
mospheric resuspension of lead-contaminated soils and road dust is unperturbed. We
estimate the following model:
Prob Yijt 51jDj,Et,Mi,Ai,Zt,St,Fj,Hj,Pj,Ij
5L½aij 1β1Dj1β2Et1dDj×Et
1G1Mi1G2Ai
1G3Zt1G4St1λ1Fj1λ2Hj1λ3Pj1λ4Ij1zj:
(2)
The definition of terms carries over from equation (1). The term Etequals 1 if a child’s
blood was drawn during the episode of depressed avgas sales from 9/2001 to 11/2001.
The impact of the deposition shock is captured by the coefficient β
2
. The coefficient of
interaction (d) measures the combined effect of airport proximity (D
j
) and the episode
indicator (E
t
). To the extent that child BLL is linked to avgas deposition, β
2
should be
negative and dpositive, the latter expectation reflecting the dissipation of the shock
effect in distance.
While the above test works to identify a distance effect, it imprecisely assumes that
PEA traffic is the same across airports and that prevailing winds behave uniformly at
each airport. For a subset of all 27 airports inventoried in the FAAOP system, we ob-
tained data on the monthly flow of PEA traffic as well as prevailing wind direction
distribution data. We use these data to analyze whether child BLLs increase with
the volume of PEA traffic and downwind risk. Importantly, the next tests exploit var-
iation in PEA traffic determined by exogenous fluctuations in local weather condi-
tions.
10
They also leverage prevailing wind direction as an exogenous source of varia-
10. Local conditions vary meaningfully across airport facilities examined. The average annual
number of snow days and precipitation inches varies considerably across airports. For instance,
CIU (in the northeast end of Michigan’s Upper Peninsula) has more than twice the number of
average annual snow days as DET (that is 9 km northeast of Detroit’s central business district).
Not only does total precipitation vary across examined airports, but so does the peak month of
precipitation and the percentage difference between peak and trough months over the calendar
year. Variation in precipitation across airports, and within airports in time, importantly deter-
mine the level of PEA traffic and consequent deposition of lead on neighborhoods nearby.
dummy variables for the year of observation (with 2001 as our reference year) and ε
i
is our error
term. Results on our shock variable indicate that avgas sales declined significantly in the after-
math of 9/11 (b15–0:57, p5:015). This result is corroborated by a fixed effects least square
model of PEA traffic in our subset of FAAOP airports. Clustering error at the airport level,
we estimated the following: ln(Pij)5β01β1shock 1G1month 1G2year 1G3airportj1εij ,
where P
ij
is the volume of PEA operations (departures and arrivals) in month iat FAAOP air-
port j,shock,month, and year are the same as before, and airport is a suite of dummy variables
corresponding to each FAAOP airport (with APN, Alpena, as our reference airport). In the
months following 9/11, PEA traffic declined significantly (b15–0:14, p5:011).
Effect of Leaded Aviation Gasoline on Blood Lead in Children Zahran et al. 585
This content downloaded from 129.082.187.225 on April 12, 2017 14:16:26 PM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
tion with respect to the problem of residential selection. The augmented regression
model is:
Prob Yijt 51jDj,Tjt,Wj,Mi,Ai,Zt,St,Fj,Hj,Pj,Ij
5L½aij 1β1Dj1β2Tjt 1β3Wj1G1Mi1G2Ai
1G3Zt1G4St1λ1Fj1λ2Hj1λ3Pj1λ4Ij1zj:
(3)
All terms carry over from equation (1), with the addition of T
jt
representing the
monthly sum of PEA arrivals and departures at the nearest airport corresponding
to the month of child blood draw, and W
j
denoting the percentage of prevailing wind
days that drift in the direction of child’s residential location.
Finally, we analyze various two-way and three-way interactions of the main treat-
ment variables. For instance, we estimate how the PEA traffic effect (T) varies by dis-
tance (D) and downwind risk (W). The expectation is that, to the extent that PEA
traffic is an important source of elevated BLL risk, the PEA traffic effect should atten-
uate in distance and amplify in downwind days. We estimate the following:
Prob Yijt 51jDj,Tjt,Wj,Mi,Ai,Zt,St,Fj,Hj,Pj,Ij
5L½aij 1β1Dj1β2Tjt 1β3Wj1dDj×Tjt
1JWj×Tjt
1ϑDj×Wj×Tjt
1G1Mi1G2Ai
1G3Zt1G4St1λ1Fj1λ2Hj1λ3Pj1λ4Ij1zj:
(4)
The interaction between PEA traffic and tract distance is captured by d.
11
The effect
of PEA traffic on downwind risk is captured by J. Moreover, the three-way effect of
distance, traffic, and downwind risk is captured by ϑ. Here, we expect the risk of el-
evated blood lead from PEA traffic(T) to amplify in downwind days (W) and to dis-
sipate in distance (D).
3. RESULTS
Table 2 reports descriptive statistics on mean BLLs and the proportion of observed
children exceeding present and past CDC reference values. All covariates behave as
expected. The proportion of children with BLL above relevant thresholds increases
11. This test also addresses a modest sampling gradient in distance to airports. Children re-
siding near airports are slightly more likely to have their blood sampled for lead content. The
sampling ratio increases less than 1% (b5–0:86, 95% CI: –1.13, –0.58) for every kilometer in
distance from the nearest airport, equal to about 9 fewer children sampled per kilometer of dis-
tance.
586 Journal of the Association of Environmental and Resource Economists June 2017
This content downloaded from 129.082.187.225 on April 12, 2017 14:16:26 PM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
in proximity to the nearest GNIS airport, in the monthly flow of PEA traffic, in the
percentage of wind days that drift in the direction of child’s residential neighborhood,
in the percentage of housing built before 1950, in summer and fall relative to spring
and winter, in proximity to lead-emitting TRI facilities, and in neighborhood popu-
lation density.
Table 3 reports coefficients predicting child BLLs (measured continuously) and
odds of a child’s BLL exceeding present (≥5mg/dL) and past (≥10 mg/dL) CDC ref-
erence values. Columns 1 and 4 report results from random intercept least squares
models with child BLL measured continuously. In column 4, and as compared to chil-
dren at >4 km from a GNIS airport, we find that the BLLs of children residing <1 km,
1–2 km, and 2–3 km of a GNIS airport are 5.7%, 2.9%, and 2.4% higher, respectively.
By exponentiation of the coefficient in column 2 (e
–0.025
), we find that the risk of a
child eclipsing the CDC reference value of 5 mg/dL decreases by 2.5% (95% CI: –3.4,
–1.5) for every kilometer from a GNIS airport. Similarly, in column 3, we find that
a 1 km increase in neighborhood distance from a GNIS airport reduces the odds of a
child’s BLL exceeding 10 mg/dL by a multiplicative factor of 0.971 (e
–0.03
). Columns 5
and 6 divide airport distance (D) into discrete categories (D≤1 km; 1 km >D<2 km;
2 km >D<3 km; and 3 km >D<4 km; and D>4 km).
In column 5, as compared to children residing >4 km from a GNIS airport, chil-
dren at <1 km, 1–2 km, and 2–3 km are 25.2% (e
0.225
), 16.5% (e
0.153
), and 9.1% (e
0.087
)
more likely to present with a BLL reading ≥5mg/dL, respectively. Similarly, the odds
of eclipsing the CDCs past threshold of concern (≥10 mg/dL) increases in airport prox-
imity. As shown in column 6, and as compared to children residing >4kmfroma
GNIS airport, children at <1 km and 1–2 km are 44.9% (e
0.371
) and 24.9% (e
0.222
)
more likely to supersede a BLL of ≥10 mg/dL, respectively. Thus, consistent with Mi-
randa et al. (2011), child BLLs decrease dose-responsively in distance, with the risk of
elevated blood lead leveling at approximately 2–3 km from the nearest GNIS airport.
Before proceeding, it is worth noting the intuitive behavior of other variables known
to influence BLL outcomes. For instance, in column 2, a unit increase in the percent-
age of housing stock built prior to 1950—a common proxy for the risk of Pb-based
paint exposure—increases the child’s odds of eclipsing the CDCs current threshold
of 5 mg/dL by a factor of 1.019 (95% CI: 1.018, 1.021). The econometric model also
detects the known seasonality in child BLL (Zahran et al. 2013), showing that, as com-
pared to the reference seasons of winter/spring, children having their blood drawn in
summer (e0:313 51:37) and fall (e0:220 51:25) months have significantly higher odds
of having BLL ≥5mg/dL.
Table 4 reports results from our quasi-experiment leveraging the fall off in air traf-
fic after 9/11, 2001. We analyze the likelihood of a child’s BLL eclipsing various
thresholds (including 3, 5, and 10 mg/dL) as well the response of child BLL measured
continuously. The coefficients of interest are our treatment period term capturing the
BLL of children sampled in the avgas deposition shock period following 9/11, and our
Effect of Leaded Aviation Gasoline on Blood Lead in Children Zahran et al. 587
This content downloaded from 129.082.187.225 on April 12, 2017 14:16:26 PM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
Table 2. Proportion of Children Eclipsing 5 and 10mg/dL and Mean Blood Lead by Covariates
Proportion
≥5mg/dL
Proportion
≥10 mg/dL
Mean
mg/dL
Two-Sample
t-Statistic
Distance to GNIS airport (km):
>P.50 .144 .024 2.82 –69.60
<P.50 .191 .038 3.23
Distance to FAAOP airport (km):
>P.50 .177 .036 3.14 –55.37
<P.50 .252 .055 3.76
Piston engine aircraft:
>P.50 .227 .053 3.58 21.74
<P.50 .203 .039 3.33
Downwind (%):
>P.50 .231 .050 3.57 21.17
<P.50 .199 .042 3.33
Sex:
Male .167 .031 3.02 17.81
Female .156 .028 2.92
Age of child:
<1 year .097 .012 2.38 –75.43
1 year .140 .024 2.79 –48.71
2 years .199 .040 3.33 52.67
3 years .187 .035 3.19 33.31
4 years .163 .029 2.97 –1.38
5 years .171 .034 3.02 4.00
% housing built <1950:
>P.50 .255 .053 3.77 2.9e102
<P.50 .069 .007 2.19
Season:
Winter .147 .026 2.85 –23.91
Spring .146 .024 2.82 –31.56
Summer .176 .035 3.11 29.21
Fall .171 .032 3.05 16.67
Population density:
>P.50 .233 .048 3.59 2.2e102
<P.50 .091 .012 2.37
% public assistance:
>P.50 .253 .052 3.76 2.8e102
<P.50 .072 .008 2.20
% high school1:
>P.50 .090 .011 2.36 –221.70
<P.50 .235 .048 3.60
This content downloaded from 129.082.187.225 on April 12, 2017 14:16:26 PM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
difference-in-differences term, which captures the interaction between airport proxim-
ity and treatment period. Analyses are limited to the years 2001 to 2003, and the
months of June to December.
12
Column 1 shows that children sampled in the deposi-
tion shock period had BLLs 4.9% lower (95% CI: –0.064, –0.034) than children sam-
pled outside the deposition shock period. As expected, the effect fades significantly with
distance to airport.
Similarly, column 2 shows that the odds of eclipsing 3 mg/dL declined by 13.5%
(1 –e–0:145) among children sampled in the treatment period, with the effect declining
~2.6% for every 1 km in distance to airport. Column 3 shows the risk of child BLL
exceeding the CDCs current reference value. Other things held equal, children sam-
pled in the deposition shock period had significantly lower odds of presenting a BLL
reading ≥5mg/dL. On average, treated children experienced an 11% (1 –e–0:117 ) de-
crease (95% CI: –16.5%, –5.3%) in the probability of elevated BLL (≥5mg/dL). While
this beneficial effect appears to fade with distance from the nearest GNIS airport (b5
0:005), the interaction coefficient is not statistically significant.
While the results in tables 3 and 4 corroborate and extend Miranda et al. (2011),
they do not account for the role of PEA traffic volume in determining the relationship
12. We thank two anonymous reviewers for suggesting we narrow the time window to limit
temporal and seasonal confounding.
Table 2 (Continued)
Proportion
≥5mg/dL
Proportion
≥10 mg/dL
Mean
mg/dL
Two-Sample
t-Statistic
Median home price:
>P.50 .072 .008 2.20 –282.62
<P.50 .252 .051 3.75
Pb facility <2 km:
Yes .226 .046 3.52 1.5e102
No .120 .019 2.62
Year:
>2005 .114 .017 2.64 –146.82
<2005 .216 .044 3.58
Note. Blood lead outcomes by sex, age of child, % housing built <1950, season of blood draw, popu-
lation density, % public assistance, % high school1, median home price, Pb facility count, and year are from
all observed children (N51,023,672) in Michigan from 2001 to 2009, whereas blood lead outcomes by
distance to FAAOP airport, PEA traffic, and downwind (%) are from children (N5364,292) for our sub-
set of 27 FAAOP airports.
Effect of Leaded Aviation Gasoline on Blood Lead in Children Zahran et al. 589
This content downloaded from 129.082.187.225 on April 12, 2017 14:16:26 PM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
Table 3. Random Intercept Logistic and Generalized Least Squares Coefficients Predicting Elevated BL (≥5 and 10 mg/dL) and BLLs in Children
in Michigan Residing <10 km from an Airport
ln (mg/dL)
(1)
≥5mg/dL
(2)
≥10 mg/dL
(3)
ln (mg/dL)
(4)
≥5mg/dL
(5)
≥10 mg/dL
(6)
Distance to airport –.007*** –.025*** –.030***
(.001) (.005) (.009)
Reference 5distance ≥4 km:
<1km .057*** .225*** .371***
(.018) (.067) (.120)
1–2km .029*** .153*** .222***
(.009) (.037) (.080)
2–3km .024*** .087*** .022
(.009) (.033) (.061)
3–4km .012 .053 –.026
(.008) (.033) (.058)
Reference 5age <1:
Age 1 .174*** .709*** 1.044*** .174*** .709*** 1.044***
(.003) (.021) (.047) (.003) (.021) (.047)
Age 2 .265*** .993*** 1.367*** .265*** .993*** 1.367***
(.003) (.022) (.048) (.003) (.022) (.048)
Age 3 .192*** .743*** 1.051*** .192*** .743*** 1.051***
(.003) (.022) (.046) (.003) (.022) (.046)
Age 4 .130*** .539*** .820*** .130*** .539*** .820***
(.003) (.022) (.048) (.003) (.022) (.048)
Age 5 .093*** .481*** .865*** .0934*** .481*** .865***
(.003) (.025) (.042) (.003) (.024) (.042)
590
This content downloaded from 129.082.187.225 on April 12, 2017 14:16:26 PM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
Male .030*** .116*** .132*** .030*** .116*** .132***
(.001) (.006) (.012) (.001) (.006) (.012)
Reference 5winter/spring:
Summer season .078*** .313*** .439*** .078*** .313*** .439***
(.002) (.009) (.016) (.002) (.009) (.016)
Fall season .059*** .220*** .293*** .059*** .220*** .293***
(.002) (.009) (.016) (.002) (.009) (.016)
% housing built <1950 .005*** .019*** .026*** .005*** .019*** .025***
(.000) (.001) (.001) (.000) (.001) (.001)
Population density .024*** .025 .127*** .024*** .031* .137***
(.004) (.017) (.025) (.004) (.017) (.025)
% public assistance .0241*** .071*** .082*** .024*** .071*** .083***
(.001) (.004) (.005) (.001) (.004) (.005)
%≥high school education –.002*** –.007*** –.008*** –.002*** –.007*** –.008***
(.000) (.002) (.002) (.000) (.002) (.002)
Median home price ($10,000) –.005*** –.021*** –.010** –.004*** –.021*** –.010*
(.001) (.003) (.006) (.001) (.003) (.006)
Pb facility <2 km .005 .007 .013 .005** .007 .014
(.003) (.011) (.015) (.003) (.011) (.015)
Reference 5year 2001:
Year 2002 –.093*** –.266*** –.310*** –.093*** –.266*** –.310***
(.003) (.014) (.027) (.003) (.012) (.027)
Year 2003 –.215*** –.591*** –.641*** –.215*** –.591*** –.641***
(.003) (.015) (.028) (.003) (.015) (.028)
Year 2004 –.254*** –.600*** –.741*** –.254*** –.600*** –.741***
(.003) (.016) (.030) (.003) (.016) (.030)
591
This content downloaded from 129.082.187.225 on April 12, 2017 14:16:26 PM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
Table 3 (Continued)
ln (mg/dL)
(1)
≥5mg/dL
(2)
≥10 mg/dL
(3)
ln (mg/dL)
(4)
≥5mg/dL
(5)
≥10 mg/dL
(6)
Year 2005 –.243*** –.628*** –.674*** –.243*** –.628*** –.674***
(.003) (.019) (.031) (.003) (.019) (.031)
Year 2006 –.300*** –.854*** –1.021****** –.300*** –.854*** –1.021***
(.003) (.019) (.033) (.003) (.019) (.033)
Year 2007 –.323*** –.906*** –1.069*** –.323*** –.906*** –1.069***
(.003) (.021) (.035) (.003) (.021) (.035)
Year 2008 –.392*** –1.212*** –1.302*** –.392*** –1.212*** –1.302***
(.003) (.019) (.036) (.003) (.019) (.036)
Year 2009 –.388*** –1.350*** –1.498*** –.388*** –1.350*** –1.498***
(.003) (.021) (.038) (.003) (.021) (.038)
Constant .747*** –2.591*** –5.541*** .710*** –2.734*** –5.672***
(.040) (.120) (.186) (.031) (.126) (.186)
Log pseudolikelihood –378,483.29 –114,715.91 –378,480.7 –114,708.21
Wald χ
2
10,219.15 12,501.98 7,020.23 54,572.13 12,504.77 7,023.79
N1,023,672 1,023,672 1,023,672 1,023,672 1,023,672 1,023,672
Number of tracts 2,431 2,431 2,431 2,431 2,431 2,431
Note. Robust standard errors clustered by census tract in parentheses. All children surveilled in Michigan Department of Community Health data system from 2001 to
2009 that reside within 10 km of a GNIS airport are included in the analysis. Distance is measured in kilometers from the population-weighted centroid of each census tract
where a child resides to the nearest GNIS airport. The blood lead thresholds of ≥5mg/dL and ≥10 mg/dL correspond to the CDCs current and past reference values for
elevated blood lead, respectively.
*p<.10.
** p<.05.
*** p<.01.
592
This content downloaded from 129.082.187.225 on April 12, 2017 14:16:26 PM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
Table 4. Difference-in-Differences Random Intercept Logistic and Generalized Least
Squares Coefficients Predicting Elevated BL (≥3, 5, and 10 mg/dL) and BLLs in Children
in Michigan Residing <10 km from an Airport
ln (mg/dL)
(1)
≥3mg/dL
(2)
≥5mg/dL
(3)
≥10 mg/dL
(4)
Distance to airport –.008*** –.026*** –.025*** –.020
(.002) (.008) (.008) (.014)
Treatment period –.049*** –.145*** –.117*** –.048
(.008) (.031) (.032) (.053)
Distance to airport ×Treatment period .005** .025** .005 –.013
(.002) (.011) (.010) (.017)
Reference 5age <1:
Age 1 .259*** .708*** .835*** 1.180***
(.007) (.031) (.038) (.070)
Age 2 .372*** 1.032*** 1.106*** 1.432***
(.008) (.032) (.038) (.074)
Age 3 .259*** .747*** .771*** .965***
(.008) (.032) (.039) (.072)
Age 4 .187*** .530*** .578*** .801***
(.007) (.033) (.039) (.070)
Age 5 .152*** .386*** .566*** .873***
(.009) (.037) (.042) (.075)
Male .034*** .100*** .107*** .127***
(.003) (.012) (.014) (.023)
Reference 5winter/spring:
Summer season .111*** .275*** .411*** .549***
(.006) (.026) (.032) (.051)
Fall season .090*** .243*** .314*** .380***
(.007) (.026) (.032) (.051)
% housing built <1950 .006*** .017*** .020*** .027***
(.000) (.001) (.001) (.002)
Population density .033*** .066*** .122*** .154***
(.006) (.024) (.023) (.034)
% public assistance .027*** .068*** .076*** .088***
(.002) (.005) (.005) (.006)
%≥high school education –.002*** –.003 –.007*** –.011***
(.001) (.002) (.002) (.003)
Median home price ($10,000) –.005*** –.030*** –.023*** –.002
(.001) (.004) (.006) (.010)
Pb facility <2 km .011*** .040** .019 .019
(.004) (.017) (.015) (.021)
593
This content downloaded from 129.082.187.225 on April 12, 2017 14:16:26 PM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
between airport distance and lead exposure risk. A more telling test would evaluate
BLL levels in response to PEA traffic. We begin with an ecological view of the data.
Figure 2 (panel A) shows the joint movement of monthly average BLL over all
measured children in Michigan (residing <10 km from 27 airports with valid PEA
traffic), as well as the average monthly sum of PEA departures and arrivals (at the
same 27 airports). Both series are standardized (m50, j51). The series share
strikingly similar seasonality and drift downward together in time. The temporal cor-
relation is strong (r50:823). While figure 2, panel A, is strongly suggestive of an
avgas and BLL link, recall that soil resuspension is a known source of seasonal varia-
tion in child BLLs (Zahran et al. 2013). Panel B addresses this potential confounding.
Again, time is on the x-axis, but now monthly average BLL is divided into two cate-
gories: above average and below average PEA traffic. The series diverge intuitively with
the high traffic series lying strictly above the low traffic series.
Narrowing in, table 5 reports coefficients that predict the likelihood of threshold
exceedance as a function of PEA traffic and wind direction. The population is restricted
to children residing within 10 km of an FAAOP airport (with valid monthly PEA traf-
fic). Recall, to estimate the effect of PEA traffic, children are matched spatially to the
Table 4 (Continued)
ln (mg/dL)
(1)
≥3mg/dL
(2)
≥5mg/dL
(3)
≥10 mg/dL
(4)
Reference 5year 2001:
Year 2002 –.126*** –.386*** –.364*** –.315***
(.006) (.025) (.024) (.042)
Year 2003 –.233*** –.687*** –.684*** –.650***
(.006) (.028) (.025) (.043)
Constant .528*** –1.120*** –2.951*** –5.767***
(.046) (.173) (.178) (.256)
Log pseudolikelihood –78,307.92 –66,301.19 –27,316.44
Wald χ
2
10687.35 4,293.10 5,325.21 2,603.60
N139,802 139,802 139,802 139,802
Number of tracts 2,420 2,420 2,420 2,420
Note. Robust standard errors clustered by census tract in parentheses. All children who reside within
10 km of a GNIS airport (N5448) and sampled from 2001 to 2003, and in the months of June to De-
cember are included in the analysis. The treatment period is September to November in 2001 correspond-
ing to measurable downward shocks in avgas sales and piston-engine aircraft traffic in Michigan following
9/11 (see n. 11).
*p<.10.
** p<.05.
*** p<.01.
594 Journal of the Association of Environmental and Resource Economists June 2017
This content downloaded from 129.082.187.225 on April 12, 2017 14:16:26 PM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
Figure 2. Monthly blood Pb (of children ≤10 km of traffic airport) and piston engine aircraft
traffic in time and blood lead levels by PEA traffic. Average monthly blood lead data correspond
to all children in Michigan residing within 10 km of a FAAOP airport with valid PEA traffic
data. Both the monthly blood lead and PEA traffic series are z-score standardized with mean 5
0, and standard deviation 51.
This content downloaded from 129.082.187.225 on April 12, 2017 14:16:26 PM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
Table 5. Random Intercept Logistic Coefficients Predicting Elevated BL (≥5 and 10 mg/dL) and BLLs in Children in Michigan Residing <10 km
from an Airport with Valid Piston Engine Aircraft Traffic
≥5mg/dL
(1)
≥10 mg/dL
(2)
≥5mg/dL
(3)
≥10 mg/dL
(4)
≥5mg/dL
(5)
≥10 mg/dL
(6)
Distance to airport –.035*** –.035** –.030*** –.030** –.030*** –.030**
(.009) (.014) (.010) (.014) (.010) (.014)
PEA traffic .028*** .035*** .034*** .038*** .033*** .038***
(.006) (.010) (.006) (.010) (.006) (.010)
Downwind .018*** .019*** .016*** .019*** .014*** .017***
(.005) (.007) (.005) (.007) (.005) (.007)
Distance ×PEA traffic–.011*** –.010*** –.011*** –.010***
(.002) (.003) (.002) (.003)
Downwind ×PEA traffic .003** –.000 .002** –.000
(.001) (.002) (.001) (.002)
Distance ×Downwind .005** .004
(.002) (.003)
Distance ×PEA traffic×Downwind .001** .000
(.000) (.001)
Reference 5age <1:
Age 1 .894*** 1.230*** .895*** 1.231*** .895*** 1.231***
(.029) (.061) (.029) (.061) (.029) (.061)
596
This content downloaded from 129.082.187.225 on April 12, 2017 14:16:26 PM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
Age 2 1.223*** 1.598*** 1.225*** 1.599*** 1.225*** 1.599***
(.028) (.060) (.028) (.060) (.029) (.06)
Age 3 .983*** 1.275*** .985*** 1.276*** .985*** 1.276***
(.028) (.058) (.028) (.058) (.028) (.058)
Age 4 .763*** 1.023*** .763*** 1.024*** .764*** 1.024***
(.028) (.059) (.028) (.059) (.028) (.059)
Age 5 .692*** 1.039*** .694*** 1.040*** .694*** 1.040***
(.031) (.061) (.031) (.061) (.031) (.061)
Male .117*** .131*** .117*** .132*** .117*** .131***
(.010) (.017) (.010) (.017) (.010) (.017)
Reference 5winter/spring:
Summer season .295*** .403*** .284*** .394*** .282*** .393***
(.014) (.023) (.012) (.023) (.014) (.023)
Fall season .204*** .278*** .198*** .274*** .198*** .274***
(.012) (.021) (.013) (.021) (.013) (.021)
% housing built <1950 .020*** .028*** .020*** .027*** .020*** .027***
(.001) (.002) (.001) (.002) (.001) (.002)
Population density .130*** .137*** .132*** .134*** .128** .135***
(.029) (.041) (.028) (.041) (.028) (.041)
% public assistance .073*** .085*** .073*** .085*** .072*** .084***
(.006) (.007) (.006) (.008) (.006) (.007)
Median home price ($10,000) –.017*** –.001 –.017*** –.002 –.017*** –.002
(.005) (.008) (.005) (.008) (.005) (.008)
597
This content downloaded from 129.082.187.225 on April 12, 2017 14:16:26 PM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
Table 5 (Continued)
≥5mg/dL
(1)
≥10 mg/dL
(2)
≥5mg/dL
(3)
≥10 mg/dL
(4)
≥5mg/dL
(5)
≥10 mg/dL
(6)
%≥high school education –.009*** –.014*** –.009*** –.015*** –.010*** –.015***
(.002) (.003) (.002) (.003) (.002) (.003)
Pb facility <2 km .012 .011 .014 .019 .021 .018
(.016) (.024) (.017) (.024) (.017) (.025)
Reference 5year 2001:
Year 2002 –.207*** –.170*** –.205*** –.167*** –.203*** –.166***
(.020) (.032) (.020) (.032) (.020) (.032)
Year 2003 –.507*** –.530*** –.500*** –.525*** –.497*** –.523***
(.022) (.037) (.023) (.037) (.023) (.037)
Year 2004 –.542*** –.605*** –.535*** –.599*** –.532*** –.598***
(.024) (.037) (.024) (.037) (.024) (.038)
Year 2005 –.510*** –.490*** –.494*** –.477*** –.488*** –.475***
(.025) (.041) (.026) (.042) (.026) (.042)
Year 2006 –.720*** –.856*** –.700*** –.841*** –.694*** –.839***
(.025) (.046) (.026) (.047) (.026) (.047)
598
This content downloaded from 129.082.187.225 on April 12, 2017 14:16:26 PM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
Year 2007 –.771*** –.928*** –.748*** –.910*** –.742*** –.908***
(.028) (.052) (.029) (.052) (.029) (.053)
Year 2008 –1.120*** –1.179*** –1.099*** –1.162*** –1.093*** –1.160***
(.029) (.052) (.029) (.052) (.029) (.053)
Year 2009 –1.232*** –1.371*** –1.207*** –1.352*** –1.201*** –1.350***
(.035) (.059) (.035) (.059) (.035) (.059)
Constant –2.984*** –5.738*** –2.849*** –5.554*** –2.800*** –5.522***
(.231) (.304) (.212) (.280) (.211) (.280)
Log pseudolikelihood –152,450.56 –55,391.70 –152,416.68 –55,383.64 –152,405.51 –55,383.67
Wald χ
2
6,603.20 4,111.73 6,868.49 4,309.68 6,985.19 4,319.15
N364,292 364,292 364,292 364,292 364,292 364,292
Number of tracts 745 745 745 745 745 745
Note. Robust standard errors clustered by census tract in parentheses. All children surveilled in Michigan Department of Community Health data system from 2001 to 2009
who reside within 10 km of a FAAOP airport (N527) are included in the analysis. Distance is measured in kilometers from the population-weighted centroid of each census
tract where a child resides to the nearest FAAOP airport. PEA traffic corresponds to the observed number of departures and arrivals at a FAAOP airport matched to the month
of blood draw for a potentially exposed child. The blood lead thresholds of ≥5mg/dL and ≥10 mg/dL correspond to the CDCs current and past reference values for elevated blood
lead, respectively. In columns 3–6, interacted variables are centered at their means.
*p<.10.
** p<.05.
*** p<.01.
599
This content downloaded from 129.082.187.225 on April 12, 2017 14:16:26 PM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
nearest FAAOP airport and temporally to the month of blood draw.
13
The reported
test exploits variation in lead deposition from PEA traffic that is at least partly gov-
erned by exogenous local meteorological conditions. These conditions vary meaning-
fully across FAAOP airport locations.
Columns 1 and 2 report main effects for the distance, PEA traffic volume, and
downwind risk variables. A 1 km increase in airport distance decreases the odds of a
child eclipsing both present and past CDC thresholds by 3.4%. These distance effects
for our subset of FAAOP airports are consistent with the distance effects reported in
table 3 for all GNIS airports. Staying with columns 1 and 2, we find that an increase of
100 PEA operations per month increases the odds that a child’s BLL ≥5mg/dL by a
factor of 1.028 (95% CI: 1.019, 1.040), and by a factor of 1.036 (95% CI: 1.017, 1.056)
with respect to the odds of a child’s BLL ≥10 mg/dL.
Columns 3 and 4 in table 5 report coefficients on the risk of elevated BLL among
sampled children for two-way interactions involving PEA traffic. Intuitively, the PEA
traffic exposure effect decreases in distance and increases in the percentage of down-
wind days. Thus, PEA traffic affects children proximate to airports more strongly than
children distant from airports. The increased likelihood of exceeding 5 mg/dL for a
given increase in PEA traffic of 100 operations decreases about 1% for every 1 km in-
crease in airport distance. Regarding the interaction of PEA traffic and downwind
risk, column 3 shows that the PEA traffic exposure effect increases a third of a percent
for every 1% increase in downwind days. Columns 5 and 6 show coefficients for the
three-way interaction of the main risk variables. As expected, prevailing wind direction
functions to attenuate the airport proximity effect of PEA traffic. Given the positive
coefficient on the two-way interaction of distance and downwind risk, the three-way
interaction can be interpreted as showing that prevailing wind expands the radius of
at-risk children.
Figure 3, panels A and B, plots results from column 3 in table 5. In both panels, the
predicted probability of a child’s BLL level ≥5mg/dL is on the y-axis and PEA trafficis
on the x-axis (moving in percentile rank units). Panel A summarizes the effect of PEA
traffic at three distances (1 km, 4 km, and 7 km) from the nearest FAAOP airport.
Predicted probabilities are derived with all other covariates fixed at their sample means.
At 7 km from a FAAOP airport, change in PEA traffic has no meaningful effect on the
13. Results involving other operationalizations of PEA traffic exposure risk, including cur-
rent month, previous month, 2 months previous, as well as 2- and 3-month rolling averages of
PEA traffic are available from the authors on request. Including more than one operation of
PEA traffic produces severe multicollinearity. The current month versus prior month PEA traf-
fic correlation is very high (r50:955). Consistent with the known half-life of lead in the blood
stream, we find that the PEA traffic effect dissipates in time lag. The 1-month lag coefficient is
half the size of the current month PEA operationalization, with the effect of the 2-month lag on
the likelihood of elevated blood lead being indistinguishable from chance.
600 Journal of the Association of Environmental and Resource Economists June 2017
This content downloaded from 129.082.187.225 on April 12, 2017 14:16:26 PM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
Figure 3. Predicted probabilities of elevated blood Pb (≥5mg/dL) by PEA traffic and distance
to nearest airport and by PEA traffic and downwind risk. Panels A and B graph results from
column 3 in table 5. In both panels, the predicted probability of a child’s BLL level ≥5mg/dL
is on the y-axis, and PEA traffic in on the x-axis (moving in percentile rank units). Panel A sum-
marizes the effect of PEA traffic at three distances (1 km, 4 km, and 7 km) from the nearest
FAAOP airport. Predicted probabilities are derived with all other covariates fixed at their sam-
ple means.
This content downloaded from 129.082.187.225 on April 12, 2017 14:16:26 PM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
BLLs of children. At 4 km from an airport, PEA traffic has a modest effect on the pre-
dicted probability of a child clearing the CDC’s threshold of concern (≥5mg/dL) when
going from the 5th to the 95th percentile in PEA traffic. At 1 km, changing PEA traffic
has a pronounced effect, increasing the predicted probability of threshold exceedance in
going from the 5th to the 95th percentile in PEA traffic by about 70%. Thus, as ex-
pected, the PEA traffic effect amplifies in airport proximity.
Panel B summarizes the effect of PEA traffic at three downwind conditions—5%,
10%, and 20%—capturing the percentage of wind days that flow toward the neighbor-
hood of a child. Other things held equal, the exceedance probability with respect to
PEA traffic increases more steeply as we move from low to high downwind risk. At
the 5th percentile in PEA traffic, children exposed to 5% downwind risk have a pre-
dicted probability of elevated BLL (≥5mg/dL) of 0.169 (95% CI: 0.151, 1.186), while
children facing 20% downwind risk have a predicted probability of elevated BLL of
0.192 (95% CI: 0.179, 0.205). At the 95th percentile in PEA traffic, the differential
in the predicted probability of threshold exceedance among children at 5% versus 20%
downwind risk increases (from 0.023 to 0.058).
4. SOCIAL BENEFITS
To quantify the significance of the results for policy, we conservatively estimate the
social benefits of a reduction in monthly PEA traffic from the 50th (407) to the
10th (133) percentile in total departures and arrivals, equivalent to a two-thirds re-
duction in avgas deposition at the representative airport. Our choice to emphasize a
movement from the 50th to 10th percentile corresponds to a reduction in PEA traffic
at the representative airport to near zero, while staying within the support of the es-
timated distribution. This two-thirds reduction scenario also happens to coincide with
the fraction of the existing fleet that could transition to motor vehicle gasoline with
minimal adjustments (Kessler 2013). Marginal damage estimates behave consistently
across various reduction scenarios.
14
To estimate the social benefit of reduced avgas consumption, we leverage the re-
gression coefficients from equation (3), and we use a standard syllogism in environ-
mental health economics linking BLL to IQ point loss and IQ point loss to future
earnings (Schwartz 1994; Grosse et al. 2002; Gould 2009). Table 6 summarizes the
steps. First, according to Census Bureau data and tract distance calculations to the
nearest airport, a total of 164,782 children reside within 2 km of an airport facility
in Michigan. Columns A and B estimate the number of children falling into various
14. As discussed below, moving from the 50th to the 10th percentile implies a marginal
damage estimate of $10.69 per gallon. In contrast, moving from the 95th to the 5th percentile
implies $11.13 per gallon, 90th to 10th implies $10.95 per gallon, 75th to 25th implies $10.85
per gallon, and 25th to 10th implies $10.55 per gallon.
602 Journal of the Association of Environmental and Resource Economists June 2017
This content downloaded from 129.082.187.225 on April 12, 2017 14:16:26 PM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
Table 6. Estimated Gain in Present Discounted Value of Lifetime Earnings from IQ Point Gain from Reduction in PEA Traffic from 50th to 10th Percentile
Risk Categories
Affected
Children (No.)
under 10th
Percentile
PEA Traffic
(A)
Affected
Children (No.)
under 50th
Percentile
PEA Traffic
(B)
Average
BLL per
Risk Bin
(mg/dL)
(C)
Average
IQ Point
Loss per
mg/dL
(D)
IQ Point Loss
under 10th
Percentile
PEA Traffic
(E)
IQ Point Loss
under 50th
Percentile
PEA Traffic
(F)
IQ Point Gain
Attributable to
PEA Traffic
Decrease from 50th
to 10th Percentile
(G)
Gain in Present
Discounted Value of
Lifetime Earnings ($)
from Decrease in PEA
Traffic ($1 Million)
(H)
<5mg/dL 132,806 131,151 2.40 0 0 0 0 $0
[131,522–134,089] [129,942–132,359]
5–10 mg/
dL 25,673 26,764 6.42 .513 84,552 88,145 3,593 $64.00
[24,933–26,413] [26,067–27,460] [82,115–86,990] [85,850–90,440] [3,450–3,735] [$61.46–$66.53]
10–20 mg/
dL 5,398 5,870 13.55 .19 31,331 34,075 2,744 $48.88
[4,980–5,815] [5,460–6,281] [28,906–33,756] [31,693–36,456] [2,700–2,787] [$48.10–$49.65]
>20 mg/dL 906 997 28.84 .11 7,250 7,982 732 $13.03
[780–1,032] [896–1,099] [6,242–8,257] [7,169–8,794] [537–927] [$9.57–$16.51]
Total 164,782 164,782 123,133 130,202 7,069 $125.90
[117,263–129,003] [124,712–135,690] [6,687–7,449] [$119.11–$132.69]
Note. Estimated count of children by BLL categories in columns A and B are derived from equation (3), setting T(representing the monthly sum of PEA arrivals and departures) at
133 for 10th percentile in PEA traffic and 407 for the 50th percentile in PEA traffic, and fixing other covariates at sample means. Row 2, column E5A×C×D; row 2, column F5B×
C×D; row 2, column G5F–E; and column H5G×$17,
This content downloaded from 129.082.187.225 on April 12, 2017 14:16:26 PM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
BLL categories, ranging from <5mg/dL to >20 mg/dL under 10th and 50th percentile
levels of monthly PEA traffic respectively. These BLL categories correspond to ob-
served breaks in the nonlinear association of IQ and BLL (Lanphear et al. 2005; Gould
2009). The count of children per BLL category is estimated by equation (3) under 10th
and 50th percentile traffic scenarios.
15
The number of children above the CDC’s reference value of 5mg/dL is higher in
column B(reflecting more PEA traffic) than in column A(reflecting less PEA traffic).
Columns Cand Dindicate the average BLL level within each BLL category and the
average IQ point loss per mg/dL, respectively. The marginal effects in column Dare
from Gould (2009), Lanphear et al. (2005), and Canfield et al. (2003). Columns E
and Festimate IQ point loss under 10th and 50th percentile PEA traffic by multiply-
ing the estimated number of affected children (in cols. Aor B), the average BLL level
per at-risk category, and the average IQ point loss per mg/dL by BLL category. The
sum of IQ points gained in going from the 50th to the 10th percentile in PEA traffic
(7,069 IQ points) is reported in column G. This reflects the difference between col-
umns Fand E.
Following others (Schwartz 1994; Salkever 1995; Grosse et al. 2002; Nevin et al.
2008), each IQ point gained corresponds to a gain in the present discounted value of
lifetime earnings of $17,815 (2006 US$). Multiplying this by the sum of IQ points
gained (7,069) gives a total social benefit of $126 million (95% CI: $119–$133 mil-
lion). This benefit would be realized for subsequent cohorts of children (0–5 years of
age) in Michigan. Assuming population density near airports and other conditions in
Michigan generalize, this suggests a national benefit of about $4.9 billion.
16
It also im-
plies an external social cost of $10.69 per gallon for currently formulated avgas in
Michigan. This estimate is not comprehensive since it reflects gains to only a subset
of the population (children ≤5 years of age), and it considers only one benefit channel
(IQ loss). Including health care and special education costs averted, as well as behav-
ioral and crime control costs, would lead to a higher estimate (Gould 2009).
15. Fixing other covariates at their means, we estimate the proportion of children exceeding
specified thresholds under 10th and 50th percentile PEA traffic scenarios. The derived propor-
tions are then multiplied by the count of children in census tracts within 2 km of an airport
(specifically, 164,782) to get the count of children per BLL category.
16. In Michigan, there are 164,782 children within 2 km of airports, while the correspond-
ing national number is an estimated 6.4 million. Scaling the Michigan benefit estimate nation-
ally, under the 50th percentile in PEA traffic, total IQ point loss attributable to PEA trafficis
405,583 and social damages are $7.2 billion. Nationally, under the 10th percentile in traffic,
total IQ point loss attributable to PEA traffic is 129,740 and social damages are $2.3 billion.
The difference in social damages under 50th and 10th percentile of PEA traffic gives our figure
of $4.9 billion.
604 Journal of the Association of Environmental and Resource Economists June 2017
This content downloaded from 129.082.187.225 on April 12, 2017 14:16:26 PM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
5. CONCLUSION
The consequences of lead exposure in childhood are lasting. Neural-imaging studies
find that adults exposed to lead as children have reduced gray matter in regions of the
brain known to govern executive judgment, impulsivity and mood regulation (Cecil
et al. 2008, 2011). Economists have convincingly linked these intellectual and socio-
emotional traits of judgment and impulsivity to long-term life outcomes (Doyle et al.
2009; Cunha and Heckman 2010; Currie and Almond 2011). Consistent with this gen-
eral literature on the long reach of childhood, Jessica Reyes (2015, 1) has shown that per-
sons exposed to lead in early life experience “an unfolding series of adverse behavioral
outcomes: behavior problems as a child, pregnancy and aggression as a teen, and crim-
inal behavior as a young adult.”
Past lead control efforts have generated sizable social benefits (Grosse et al. 2002;
Gould 2009; Pichery et al. 2011; Jones 2012), with mean BLLs for children one to
five years old declining from 14.9 mg/dL in 1976 to 1.7 mg/dL two decades later
(Gould 2009). Despite this dramatic success, BLLs remain high for over half a million
children in the United States (Zahran et al. 2011). The current study provides clear
evidence that elevated BLLs in children proximate to airports is at least partly attrib-
utable to avgas deposition from piston-engine aircraft.
Specifically, the odds that a child’s BLL will eclipse CDC thresholds for concern
increases dose-responsively in proximity to airports, declines measurably in neighbor-
hoods proximate to airports in the months following 9/11, increases dose-responsively
in the flow of PEA traffic, and increases significantly in the percentage of downwind
risk days. Meanwhile, statistical interactions between residential distance, PEA traffic,
and downwind risk all behave in intuitive ways, supporting the claim that avgas depo-
sition is an independent source of lead exposure risk for children. As shown in table 3,
children residing within 1 km of a GNIS airport are 25% and 45% more likely to ex-
ceed present and past thresholds of concern than children at ≥4 km from an airport.
As shown in figure 3, panel A, the predicted probability of exceeding the current CDC
threshold for concern for a child residing within 1 km of airport nearly doubles in go-
ing from low (5th percentile) to high (95th percentile) PEA traffic.
According to the analysis, a hypothetical reduction in PEA traffic from the 50th to
the 10th percentile would generate a 5-year cohort benefit of $126 million for Mich-
igan and $4.9 billion nationwide.
17
Accompanying such a reduction, the number of
children falling below the CDC current reference threshold of 5 mg/dL would increase
17. Our nationwide 5-year cohort benefit of $4.9 billion is similar to Wolfe et al.’s (2016)
1-year cohort estimate of $1.06 billion in economic damages from elevated atmospheric lead
exposure using the Community Multi-Scale Air Quality model. Wolfe et al. (2016) note that
the monetary impacts of aviation lead emissions are similar in magnitude to noise, climate change,
and air quality degradation from all commercial operations.
Effect of Leaded Aviation Gasoline on Blood Lead in Children Zahran et al. 605
This content downloaded from 129.082.187.225 on April 12, 2017 14:16:26 PM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
by about 1,600 children in Michigan and 64,000 children nationwide. To put this in
perspective, the recent catastrophic failure of the water treatment system in Flint,
Michigan, increased the number of children with elevated BLLs by approximately
200 (Hanna-Attisha et al. 2016).
18
The comparison is imperfect since the Flint water
crisis occurred at a different time period, with a lower baseline fraction of children
with BLLs ≥5mg/dL, and because the Flint case involved explicit acts of commission.
Nevertheless, the comparison demonstrates the large scale of social damages that can
be attributed to the ongoing consumption of avgas in the United States.
Under current regulations, these damages are unpriced. An emission fee that forced
consumers to internalize these costs—a tax of approximately $10 per gallon compared
to a pump price of approximately $6
19
—would likely cause a rapid transition away from
lead-formulated avgas by the roughly two-thirds of the existing PEA fleet for which the
lead additive is noncritical (Kessler 2013). In addition, by creating strong incentives for
innovation and for the gradual turnover of the lead-dependent fleet, such a policy would
set the stage for the eventual phase out of lead from the aviation sector.
REFERENCES
Callahan, Colleen. 2010. The plane truth: Air quality impacts of airport operations and strategies for sus-
tainability: A case study of the Los Angeles World Airports. MA thesis, University of California, Los
Angeles, Urban Planning.
Campanella, Richard, and Howard W. Mielke. 2008. Human geography of New Orleans’urban soil lead
contaminated geochemical setting. Environmental Geochemistry and Health 30 (6): 531–40.
Canfield, Richard L., Charles R. Henderson, Deborah A. Cory-Slechta, Christopher Cox, Todd A. Jusko,
and Bruce P. Lanphear. 2003. Intellectual impairment in children with blood lead concentrations below
10 mg per deciliter. New England Journal of Medicine 348:1517–26.
Carr, Edward, Mark Lee, Kristen Marin, Christopher Holder, Marion Hoyer, Meredith Pedde, Rich Cook,
and Jawad Toumac. 2011. Development and evaluation of an air quality modeling approach to assess
near-field impacts of lead emissions from piston-engine aircraft operating on leaded aviation gasoline.
Atmospheric Environment 45:5795–5804.
CDC (US Centers for Disease Control and Prevention). 2012a. Low level lead exposure harms children: A
renewed call for primary prevention. Report of the Advisory Committee on Childhood Lead Poisoning
18. Hanna-Attisha et al. (2016) find that the percentage of sampled children in Flint, Mich-
igan, with elevated BLLs increased from 2.4% to 4.9% following the water source change. We
derive the estimated count of affected children by the water source change by taking the count of
children under five (8% of a population of 102,434, Census, April 1, 2010) multiplied by the
figures reported by Hanna-Attisha et al.
19. Of course, the efficient emission tax would be applied to the lead content of gasoline, so
the tax per gallon would vary for different formulations of avgas. The estimate of $10.69 applies
to an average gallon of avgas sold in Michigan over the sample period. This is equivalent to
$5.60 per gram of lead.
606 Journal of the Association of Environmental and Resource Economists June 2017
This content downloaded from 129.082.187.225 on April 12, 2017 14:16:26 PM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
Prevention of the Centers for Disease Control and Prevention, Atlanta. http://www.cdc.gov/nceh
/lead/acclpp/final_document_030712.pdf.
——— . 2012b. CDC response to Advisory Committee on childhood lead poisoning prevention recommen-
dations in “Low level Pb exposure harms children: A renewed call of primary prevention.”Centers for
Disease Control and Prevention, Atlanta. http://www.cdc.gov/nceh/lead/acclpp/cdc_response_lead
_exposure_recs.pdf.
Cecil, Kim M., Christopher J. Brubaker, Caleb M. Adler, Kim N. Dietrich, Mekibib Altaye, John C.
Egelhoff, Stephanie Wessel, Ilayaraja Elangovan, Richard Hornung, Kelly Jarvis, and Bruce P. Lanphear.
2008. Decreased brain volume in adults with childhood lead exposure. PLoS Medicine 5 (5): e112.
Cecil, Kim M., Kim N. Dietrich, Mekibib Altaye, John C. Egelhoff, Diana M. Lindquist, Christopher J.
Brubaker, and Bruce P. Lanphear. 2011. Proton magnetic resonance spectroscopy in adults with child-
hood lead exposure. Environmental Health Perspectives 119 (3): 403–8.
Cunha, Flavio, and James J. Heckman. 2010. Investing in our young people. NBER Working Paper 16201,
National Bureau of Economic Research, Cambridge, MA.
Currie, Janet, and Douglas Almond. 2011. Human capital development before age five. In Handbook of la-
bor economics 4B, ed. David Card and Orley Ashenfelter. Amsterdam: Elsevier.
DHHS (US Department of Health and Human Sciences) National Toxicology Program. 2012. Monograph
on the health effects of low-level lead.http://ntp.niehs.nih.gov/NTP/ohat/Pb/Final/MonographHealth
EffectsLowLevelPb_Final_508.pdf.
Dietrich, Kim N., Ris M. Douglas, Paul A. Succop, Omer G. Berger, and Robert L. Bornschein. 2001.
Early exposure to lead and juvenile delinquency. Neurotoxicology and Teratology 23:511–18.
Doyle, Orla, Colm P. Harmon, James J. Heckman, and Richard E. Tremblay. 2009. Investing in early hu-
man development: Timing and economic efficiency. Economics and Human Biology 7 (1): 1–6.
Ells, Steven W. 2006. Lead is still king. AOPA Pilot 49 (5): 111–15.
EPA (US Environmental Protection Agency). 2008. National emissions inventory data and documenta-
tion. Office of Air Quality Planning and Standards, US Environmental Protection Agency, Research
Triangle Park, NC. http://www.epa.gov/ttnchie1/net/2008inventory.html.
——— . 2010a. Advance notice of proposed rulemaking on lead emissions from piston-engine aircraft using
leaded aviation gasoline. Office of Transportation and Air Quality. EPA-420-F-10-013, Washington,
DC.
——— . 2010b. Calculating piston-engine aircraft airport inventories for lead for the 2008 national emis-
sions inventory. EPA-420-B-10-044, Washington, DC.
——— . 2014. Petition for reconsideration of EPA’s denial of Friends of the Earth’s October 3, 2006 pe-
tition for rulemaking seeking the regulation of lead emissions from general aviation aircraft. EPA-HQ-
OAR-2007-0294. https://www.epa.gov/aboutepa/petition-reconsideration-epas-denial-friends-earths
-october-3-2006-petition-rulemaking.
FAA (Federal Aviation Administration). 2012. Final report: Unleaded avgas findings and recommenda-
tions. Unleaded Avgas Transition Aviation Rulemaking Committee. http://www.faa.gov/regulations
_policies/rulemaking/committees/documents/media/Avgas.ARC.RR.2.17.12.pdf.
Farfel, Mark R., Anna O. Orlova, Peter S. J. Lees, Charles Rohde, Peter J. Ashley, and J. Julian Chisolm Jr.
2003. A study of urban housing demolitions as sources of lead in ambient dust: Demolition practices
and exterior dust fall. Environmental Health Perspectives 111 (9): 1228–34.
Filippelli, Gabriel M., Mark A. S. Laidlaw, Jennifer C. Latimer, and Robyn Raftis. 2005. Urban lead poi-
soning and medical geology: An unfinished story. GSA Today 15 (1): 4–11.
Effect of Leaded Aviation Gasoline on Blood Lead in Children Zahran et al. 607
This content downloaded from 129.082.187.225 on April 12, 2017 14:16:26 PM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
Gould, Elise. 2009. Childhood lead poisoning: Conservative estimates of the social and economic benefits of
lead hazard control. Environmental Health Perspectives 117 (7): 1162–67.
Grosse, Scott D., Thomas D. Matte, Joel Schwartz, and Richard J. Jackson. 2002. Economic gains resulting
from the reduction in children’s exposure to lead in the United States. Environmental Health Perspectives
110 (6): 563–69.
Hanna-Attisha, Mona, Jenny LaChance, Richard Casey Sadler, and Allison Champney Schnepp. 2016. El-
evated blood lead levels in children associated with the Flint drinking water crisis: A spatial analysis of
risk and public health response. American Journal of Public Health 106 (2): 283–90.
Jones, David J. 2012. Primary prevention and health outcomes: Treatment of residential lead-based paint
hazards and the prevalence of childhood lead poisoning. Journal of Urban Economics 71 (1): 151–64.
Jusko, Todd A., Charles R. Henderson Jr., Bruce P. Lanphear, Deborah A. Cory-Slechta, Patrick J. Par-
sons, and Richard L. Canfield. 2008. Blood lead concentrations <10 microg/dL and child intelligence at
6 years of age. Environmental Health Perspectives 116 (2): 243–48.
Kemper, Alex R., W. Clayton Bordley, and Stephen M. Downs. 1998. Cost-effectiveness analysis of lead
poisoning screening strategies following the 1997 guidelines of the Centers for Disease Control and Pre-
vention. Archives of Pediatrics and Adolescent Medicine 152 (12): 1202–8.
Kessler, Rebecca. 2013. Sunset for leaded aviation gasoline? Environmental Health Perspectives 121 (2):
A54–A57.
Laidlaw, Mark A. S., Howard W. Mielke, Gabriel M. Filippelli, David L. Johnson, and Christopher R.
Gonzales. 2005. Seasonality and children’s blood lead levels: Developing a predictive model using cli-
matic variables and blood lead data from Indianapolis, Indiana, Syracuse, New York and New Orleans,
Louisiana (USA). Environmental Health Perspectives 113 (6): 793–800.
Laidlaw, Mark A. S., Sammy Zahran, Howard W. Mielke, Mark P. Taylor, and Gabriel M. Filippelli.
2012. Re-suspension of lead contaminated urban soil as a dominant source of atmospheric lead in Bir-
mingham, Chicago, Detroit and Pittsburgh, USA. Atmospheric Environment 49:302–10.
Lanphear, Bruce P., Richard Hornung, Jane Khoury, Kimberly Yolton, Peter Baghurst, David C. Bellinger,
Richard L. Canfield, et al. 2005. Low-level environmental lead exposure and children’s intellectual func-
tion: An international pooled analysis. Environmental Health Perspectives 113:894–99.
Lidsky, Theodore I., and Jay S. Schneider. 2003. Lead neurotoxicity in children: Basic mechanisms and
clinical correlates. Brain 126 (1): 5–19.
Manton, W. I., C. R. Angle, K. L. Stanek, Y. R. Reese, and T. J. Kuehnemann. 2000. Acquisition and
retention of lead by young children. Environmental Research 82 (1): 60–80.
Mielke, Howard W., and Sammy Zahran. 2012. The urban rise and fall of air lead (Pb) and the latent surge
and retreat of societal violence. Environment International 43:48–55.
Miranda, Marie Lynn, Rebecca Anthopolos, and Douglas Hastings. 2011. A geospatial analysis of the ef-
fects of aviation gasoline on childhood blood lead levels. Environmental Health Perspectives 119 (10):
1513–16.
Miranda, Marie Lynn, Dohyeong Kim, M. Alicia Overstreet Galeano, Christopher J. Paul, Andrew P. Hull,
and S. Philip Morgan. 2007. The relationship between early childhood blood lead levels and perfor-
mance on end of grade tests. Environmental Health Perspectives 115 (8): 1242–47.
Needleman, Herbert L., and Constantine A. Gatsonis. 1990. Low-level lead exposure and the IQ of chil-
dren. A meta-analysis of modern studies. JAMA 263 (5): 673–78.
Nevin, Rick. 2006. Understanding international crime trends: The legacy of preschool lead exposure. En-
vironmental Research 104 (3): 315–36.
608 Journal of the Association of Environmental and Resource Economists June 2017
This content downloaded from 129.082.187.225 on April 12, 2017 14:16:26 PM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
Nevin, Rick, David E. Jacobs, Michael Berg, and Jonathan Cohen. 2008. Monetary benefits of preventing
childhood lead poisoning with lead-safe window replacement. Environmental Research 106 (3): 410–19.
Nigg, Joel T., Molly Nikolas, G. Mark Knottnerus, Kevin Cavanagh, and Karen Friderici. 2010. Confir-
mation and extension of association of blood lead with attention-deficit/hyperactivity disorder (ADHD)
and ADHD symptom domains at population-typical exposure levels. Journal of Child Psychology and Psy-
chiatry 51 (1): 58–65.
Piazza, B. 1999. Santa Monica Municipal Airport: A report on the generation and downwind extent of
emissions generated from aircraft and ground support operations. Environmental Protection Agency,
Washington, DC. http://yosemite.epa.gov/oar/communityassessment.nsf/.
Pichery, Céline, Martine Bellanger, Denis Zmirou-Navier, Philippe Glorennec, Philippe Hartemann, and
Philippe Grandjean. 2011. Childhood lead exposure in France: Benefit estimation and partial cost-
benefit analysis of lead hazard control. Environmental Health 10 (1): 44
Quinn, Pete. 2013. Road density as a proxy for population density in regional-scale risk modeling. Natural
Hazards 65 (3): 1227–48.
Rabinowitz, Michael B. 1991. Toxicokinetics of bone lead. Environmental Health Perspectives 91:33.
Rau, Tomas, Sergio Urzua, and Loreta Reyes. 2015. Early exposure to hazardous waste and academic
achievement: Evidence from a case of environmental negligence. Journal of the Association of Environmen-
tal and Resource Economists 2 (4): 527–63.
Raymond, Jamie, Will Wheeler, and Mary Jean Brown. 2014. Lead screening and prevalence of blood lead
levels in children aged 1–2 years: Child blood lead surveillance system, United States, 2002–2010 and
national health and nutrition examination survey, United States, 1999–2010. Morbidity and Mortality
Weekly Reports 63 (2): 36–42.
Reyes, Jessica Wolpaw. 2007. Environmental policy as social policy? The impact of childhood lead exposure
on crime. BE Journal of Economic Analysis and Policy 7(1):1–43. http://www.bepress.com/bejeap
/vol7/iss1/art51/.
——— . 2012. Lead policy and academic performance: Insights from Massachusetts. NBER Working Paper
18327, National Bureau of Economic Research, Cambridge, MA.
——— . 2015. Lead exposure and behavior: Effects on antisocial and risky behavior among children and
adolescents. Economic Inquiry 53 (3): 1580–1605.
Salkever, David S. 1995. Updated estimates of earnings benefits from reduced exposure of children to en-
vironmental lead. Environmental Research 70 (1): 1–6.
Schwartz, Joel. 1994. Societal benefits of reducing lead exposure. Environmental Research 66 (1): 105–24.
Tetra Tech, Inc. 2007. Destin Airport air sampling project executive summary. Prepared for City of Destin,
Florida. Destin, FL. http://www.docstoc.com/docs/85649998/Project-Management-Sample-Executive
-Summary.
Wolfe, Philip J., Amanda Giang, Akshay Ashok, Noelle E. Selin, and Steven R. H. Barrett. 2016. Costs of
IQ loss from leaded aviation gasoline emissions. Environmental Science and Technology 50 (17): 9026–
33.
Zahran, Sammy, Mark A. S. Laidlaw, Shawn P. McElmurry, Gabriel M. Filippelli, and Mark Taylor. 2013.
Linking source and effect: Resuspended soil lead, air lead, and children’s blood lead levels in Detroit,
Michigan. Environmental Science and Technology 47 (6): 2839–45.
Zahran, Sammy, Howard W. Mielke, Christopher R. Gonzales, Eric T. Powell, and Stephan Weiler. 2010.
New Orleans before and after Hurricanes Katrina/Rita: A quasi-experiment of the association between
soil Pb and children’s blood lead. Environmental Science and Technology 44 (12): 4433–40.
Effect of Leaded Aviation Gasoline on Blood Lead in Children Zahran et al. 609
This content downloaded from 129.082.187.225 on April 12, 2017 14:16:26 PM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
Zahran, S., H. W. Mielke, S. Weiler, K. J. Berry, and C. R. Gonzales. 2009. Children’s blood lead and
standardized test performance response as indicators of neurotoxicity in metropolitan New Orleans
elementary schools. Neurotoxicology 30 (6): 888–97.
Zahran, Sammy, Howard W. Mielke, Stephan Weiler, and Christopher R. Gonzales. 2011. Nonlinear as-
sociations between blood lead in children, age of child, and quantity of soil lead in metropolitan New
Orleans. Science of the Total Environment 409 (7): 1211–18.
610 Journal of the Association of Environmental and Resource Economists June 2017
This content downloaded from 129.082.187.225 on April 12, 2017 14:16:26 PM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
