America’s Children and the Environment, Third Edition Environments and Contaminants Criteria Air Pollutants Methods January 2018

Environments and Contaminants: Criteria Air Pollutants

Methods

Indicator

El. Percentage of children ages 0 to 17 years living in counties with pollutant concentrations
above the levels of the current air quality standards, 1999-2016.

E2. Percentage of children ages 0 to 17 years living in counties with 8-hour ozone and 24-hour
PM2.5 concentrations above the levels of air quality standards, by frequency of occurrence, 1999-
2016.

Summary

EPA's Office of Air Quality Planning and Standards (OAQPS) has set primary (health-based)
National Ambient Air Quality Standards (NAAQS) for six common air pollutants, often referred
to as criteria air pollutants (or simply criteria pollutants). Each NAAQS has an averaging period
(e.g., one-hour, eight-hour, 24-hour, three-month, annual) and a level.1 For Indicator El we
analyzed the following NAAQS: carbon monoxide, eight-hour; ozone, eight-hour; PM10, 24-
hour; PM2.5, annual; PM2.5, 24-hour; sulfur dioxide, one-hour; nitrogen dioxide, one-hour; lead,
three-month. For indicator E2 we analyzed the annual frequency with which pollutant
concentrations were above the current levels of the eight-hour ozone and 24-hour PM2.5
standards. Monitoring data are submitted by state and local environmental agencies to the
national EPA Air Quality System database. For each NAAQS standard, the criteria pollutant
concentrations are averaged over the associated averaging period. For each NAAQS, monitor,
and year, we used air quality summary statistics to determine whether there was a concentration
above the current level of the standard. For each NAAQS, county, and year, the county exceeds
the level of an air quality standard if there was a concentration greater than the level of the
standard at any of the county's monitors. In addition to presenting data for each of the criteria
pollutants separately, Indicator El presents the percentage of children living in counties in which
concentrations were above the level of a NAAQS for any criteria air pollutant (i.e., exceedance
of standard levels for one or more criteria air pollutants).

Indicator El is the total number of children ages 0 to 17 years living in counties with pollutant
concentrations above the levels of air quality standards at any time during the year, divided by
the total number of children ages 0 to 17 years in the United States. The supplementary tables for
Indicator El provide the percentages of children ages 0 to 17 years living in counties with
pollutant concentrations above the levels of air quality standards, stratified by race/ethnicity
(Table Ela) or family income (Table Elb), for the year 2016. Indicator E2 is the total number of
children ages 0 to 17 years living in counties with a given number of occurrences of

1 Each NAAQS also has a statistical form: for example the annual 4th highest daily maximum eight-hour average
value for ozone, averaged over three years, and has various data completeness requirements. These aspects of the
NAAQS are not used in these indicators.

America's Children and the Environment, Third Edition
January 2018

Page 1

-------
Environments and Contaminants: Criteria Air Pollutants

concentrations above the levels of air quality standards for short-term exposure to ozone or PM2.5
during the year, divided by the total number of children ages 0 to 17 years in the United States.

Overview of Data Files

The following files are needed to calculate these indicators:

• Annual Summary files. There is one file for each year. For these analyses we extracted
data for CO, NO2, ozone, SO2, PM2.5 (24-hour standard only) and PM10. These comma-
separated files contain the pollutant parameter name (PARAMETER NAME), pollutant
standard (POLLUTANT STANDARD), unit of measure (UNITS OF MEASURE), code
(PARAMETER CODE), Event Type (EVENT TYPE), and sample duration (SAMPLE
DURATION), the state and county FIPS codes (STATE CODE, COUNTY CODE), site
number (SITE NUM), Pollutant Occurrence Code (POC), year, maximum value (1ST
MAX VALUE), arithmetic mean (ARITHMETIC MEAN), and the number of
exceedances of the primary standard (PRIMARY VIOLATION COUNT). These files
were obtained from the website:
https://aqs.epa.gov/aqsweb/airdata/download files.html

• County air quality exceedance summary data. There is one file for each year. This file
contains the state and county names, state and county FIPS codes, and the variable
Pm25wtdamviol that indicates whether the PM2.5 annual mean NAAQS was exceeded;
i.e., whether the weighted annual mean concentration was above the level of the NAAQS
at any monitor in the county with sufficiently complete data. Although this file has
county exceedance data for other criteria pollutants, those data were not used. These files
were obtained as SAS datasets directly from EPA.11

• Lead maximum rolling three-month averages data for 1999 to 2006. This file contains the
site ID, year, and the maximum rolling three-month average lead concentration for each
year from 1999 to 2006. This file was obtained as an Excel file directly from EPA.111

• Lead maximum rolling three-month averages data for 2007 to 2012. Year 2007 from
Pb_DesignValues_20072009_Final.xls. Year 2008 from
Pb_DesignValues_20082010_FinalRevised.xlsx. Year 2009 from
Pb_DesignValues_2009201 l_FINAL_07_26_12.xlsx. Year 2010 from
Pb_DesignValues_20102012_FINAL_07_22_13 .xlsx. Year 2011 from
Pb_DesignValues_20112013_FINAL_08_13_14.xlsx. Year 2012 from
Pb_DesignValues_20122014_FINAL_08_03_15 .xlsx.

These files contain the site ID, the rolling three-month average lead concentration for
each month and year from 2007 to 2012, and the annual maximum rolling three-month
average lead concentration for each year from 2011 to 2014. The files were originally

II David Mintz, EPA OAQPS, 919- 541-5224. Mintz.David@epamail.epa.gov

III Mark Schmidt, EPA OAQPS, 919-541-2416. Schmidt.Mark@epamail.epa.gov. Current contact: Halil Cakir, EPA

America's Children and the Environment, Third Edition Page 2

January 2018

-------
Environments and Contaminants: Criteria Air Pollutants

obtained from the URL: http://epa.gov/airtrends/values previous.html. The same
information can currently be found at the URL: http://www.epa.gov/air-trends/air-
quality-design-values.

•	Lead rolling three-month averages data for 2013 to 2015. Year 2013from
Pb_DesignValues_20132015_FINAL_08_29_16.xlsx. This file contains the site ID and
the annual maximum rolling three-month average lead concentration for each year from
2013 to 2015. The files were originally obtained by clicking the link for "Lead Detailed
Information" from the URL: http://epa.gov/airtrends/values.html. The same information
can currently be found at the URL: http://www.epa.gov/air-trends/air-qualitv-design-
values.

•	Lead rolling three-month averages data for 2014 to 2016. Years 2014 to 2016 from
Pb_DesignValues_20142016_final_29junel7.xlsx. This file contains the site ID and the
annual maximum rolling three-month average lead concentration for each year from 2014
to 2016. The file was found at the Lead Design Values, 2016 link at the URL:
http://www.epa.gov/air-trends/air-qualitv-design-values.

•	Census data. This file contains the state and county FIPS codes, year, and children's
population. For 1999, we obtained this information from the U.S. Census Bureau files:

Estimates of the Population of Counties by Age and Sex: 1990-1999, August 30,
2000. The file headers were "(CO-99-9) Population Estimates for Counties by Age
and Sex: Annual Time Series July 1, 1990 to July 1, 1999."
http://www.census.gov/popest/data/counties/asrh/1990s/CQ-99-09.html.

These files give county populations by age and sex for 1990 to 1999. We summed
these populations by year and county across all ages 0 to 17 years and both sexes.

For 2000-2009, we obtained this information from the bridged-race intercensal
population files:

National Center for Health Statistics. Intercensal estimates of the resident
population of the United States for July 1, 2000-July 1, 2009, by year, county, single-
year of age (0, 1, 2,.., 85 years and over), bridged race, Hispanic origin, and sex.
Prepared under a collaborative arrangement with the U.S. Census Bureau. Available
from: http://www.cdc.gov/nchs/nvss/bridged race.htm as of October 26, 2012,
following release by the U.S. Census Bureau of the revised unbridged intercensal
estimates by 5-year age group on October 9, 2012.

The children's populations by year and county were obtained by summing across the ages
0 to 17 years inclusive.

For 2010-2016, we obtained this information from the bridged race Vintage 2016
postcensal population file:

America's Children and the Environment, Third Edition
January 2018

Page 3

-------
Environments and Contaminants: Criteria Air Pollutants

National Center for Health Statistics. Vintage 2016 postcensal estimates of the
resident population of the United States (April 1, 2010, July 1, 2010-July 1, 2016), by
year, county, single-year of age (0, 1, 2,85 years and over), bridged race, Hispanic
origin, and sex. Prepared under a collaborative arrangement with the U.S. Census
Bureau. Available from: http://www.cdc.gov/nchs/nvss/bridged_race.htm as of June
26, 2017, following release by the U.S. Census Bureau of the unbridged Vintage
2016 postcensal estimates by 5-year age group on June 22, 2017.

The children's populations by year, county, and race/ethnicity were obtained by summing
across the ages 0 to 17 inclusive.

National Ambient Air Quality Standards (NAAQS)

Table 1 lists the NAAQS for the criteria pollutants used in indicator El. The first two columns
give the pollutant and the averaging time. The third column gives the corresponding level of the
standard. The fourth column gives the value used in determining, for purposes of this indicator,
whether a concentration is above the level of a standard, and incorporates a rounding convention.
For example, the eight-hour average CO standard's level of 9 ppm has a rounding convention of
1 ppm, so that the level of the standard is exceeded if any eight-hour average CO concentration
rounded to the nearest 1 ppm exceeds 9 ppm; i.e., the unrounded eight-hour average has to equal
or exceed the target value of 9.5 ppm. A county exceeds the level of the standard if one or more
of the monitors in the county has a concentration greater than the level of the standard. Except
for PM2.5, co-located monitors are treated as separate monitors. For PM2.5, if there are several co-
located monitors, then only the monitor with the lowest Monitor Number (also known as the
Pollutant Occurrence Code, abbreviated as POC) is used. We use PM2.5 data from the lowest
POC (typically POC 1, the primary sampler) to represent an area.lv For other pollutants, we use
data from all monitor/POC combinations.

It should be noted that counties with concentrations above the levels of the standards are not
necessarily non-attainment counties for the NAAQS. A non-attainment designation is generally
based on three years of data, certain data completeness criteria, and, for short-term standards,
requires multiple daily exceedances of the NAAQS. For example, non-attainment of the carbon
monoxide eight-hour standard only uses eight-hour averages with at least six hourly values, and
occurs when there are two or more eight-hour averages above the NAAQS level. Non-attainment
of the eight-hour ozone standard occurs if the three-year average of the fourth highest daily
maximum eight-hour average exceeds the level of the standard and the data completeness criteria
are met. For a detailed description of the NAAQS attainment and non-attainment calculations,
see the NAAQS website at http://www.epa.gov/naaqs/.

1V A similar approach was used in the OAQPS report "Analyses of Particulate Matter (PM) Data for the PM NAAQS

America's Children and the Environment, Third Edition Page 4

January 2018

-------
Environments and Contaminants: Criteria Air Pollutants

Table 1: Levels of the National Ambient Air Quality Standards (NAAQS) and the Values
Used to Define Exceedances for Indicators El and E2*

Pollutant

Standard Averaging Time

Level of the

Target value to define





Standard

exceedance for indicator







calculation**

Carbon monoxide

1-hour

35 ppm

Not considered



8-hour3

9 ppm

9.5 ppm

Nitrogen dioxide

Annual

0.053 ppm

Not considered



l-hourb

100 ppb

100.5 ppb

Ozone

8-hour0

0.070 ppm

0.071 ppm

Lead

Rolling 3-month averaged

0.15 ng/1113

0.155 ng/1113

PM10

24-houre

150 ng/1113

155 ng/1113

PM25

24-hourf

35 |ig/m3

35.5 |ig/m3



Annual®

12 ng/1113

12.05 ng/1113

Sulfur dioxide

1-hour11

75 ppb

75.5 ppb

* Indicators El and E2 are calculated with reference to the current averaging time and level of the air quality
standard for all years.

Standards not used for indicator El are shown as "Not considered."
aThe carbon monoxide 8-hour standard was established in 1971 (See 59 FR 38906, August 1, 1994).
b The nitrogen dioxide 1-hour standard was established in January 2010 (75 FR 6474, February 9, 2010).

0 The ozone 8-hour standard was adopted in July 1997 (62 FR 38856, July 18, 1997). The level of the ozone 8-hour
standard was changed from 0.08 to 0.075 ppm in March 2008 (73 FR 16436, March 27, 2008). The level of the
ozone 8-hour standard was changed again from 0.075 to 0.070 ppm in October 2015 (80 FR 95292, October 26,
2015). For the 2015 ozone standard level calculations, all hourly values are 8-hour average values and are reported
to three decimal places, truncating digits beyond the third decimal place.

d The level of the lead standard was changed from 1.5 to 0.15 |ig/m3 in October 2008 (73 FR 66964, November 12,

2008). The revised standard uses the maximum rolling three-month average lead concentration.

eThe form, but not the level, of the 24-hour standard for PMio was revised in July 1997 (62 FR 38652, July 18,

1997).

fThe level of the 24-hour standard for PM2 5 was revised from 65 |ig/m3 to 35 |ig/m3 in 2006 (71 FR 61144, October
17, 2006).

gThe level of the annual standard for PM2 5 was revised from 15 |-ig/m3 to 12 |-ig/m3 in 2012 (78 FR 3086, January
15,2013).

hThe final rule for the sulfur dioxide 1-hour standard was issued in June 2010 (75 FR 35220, June 22, 2010).
Air Quality Data

Annual Summary monitoring data from the EPA Air Quality System (AQS) Data Mart for the
years 1999 to 2016 were obtained from the EPA website
https://aqs.epa.gov/aqsweb/airdata/download files.html

The Annual Summary files include the year, state and county FIPS codes, site number, parameter
(pollutant) code (PARAMETER CODE), pollutant standard (POLLUTANT STANDARD), and
pollutant occurrence code (POC). The Annual Summary files also include the sample duration
(SAMPLE DURATION), and the measurement units (UNITS OF MEASURE). The event type
(EVENT TYPE) flags whether exceptional events are included in the annual summary statistics.
For these analyses we included both the event types "No Events" and "Events Inclucded" (sic)
that denote cases where either there were no exceptional events for the year ("No Events") or

America's Children and the Environment, Third Edition
January 2018

Page 5

-------
Environments and Contaminants: Criteria Air Pollutants

where there were exceptional events that year and the analysis included data for those events
("Events Inclucded"). Also included are the annual maximum value (1ST MAX VALUE) and
the number of exceedances of the primary standard (PRIMARY VIOLATION COUNT).

The following combinations of the parameter code, pollutant standard, sample duration,
measurement unit, maximum, and arithmetic mean were used to define exceedances of the
standards used for indicator El:

•	CO 8-hour standard: PARAMETER CODE = 42101, POLLUTANT STANDARD =
"CO 8-hour 1971", SAMPLE DURATION = "8-HR RUN AVG END HOUR" or "8-HR
RUN", UNITS OF MEASURE = "Parts per million", 1ST MAX VALUE > 9.5

•	N02 1 -hour standard: Either PARAMETER CODE = 42602, POLLUTANT
STANDARD = "N02 1-hour", SAMPLE DURATION = "1 HOUR", UNITS OF
MEASURE = "Parts per million", 1ST MAX VALUE > 0.1005 Or PARAMETER
CODE = 42602, POLLUTANT STANDARD = "N02 1-hour", SAMPLE DURATION =
"1 HOUR", UNITS OF MEASURE = "Parts per billion", 1ST MAX VALUE >100.5
Ozone 8-hour standard: PARAMETER CODE = 44201, POLLUTANT STANDARD =
"Ozone 8-hour 2015", SAMPLE DURATION = "8-HR RUN AVG BEGIN HOUR" or
"8-HR RUN", UNITS OF MEASURE = "Parts per million", 1ST MAX VALUE > 0.071

•	PMio 24-hour standard: PARAMETER CODE = 81102, POLLUTANT STANDARD =
"PM10 24-hour 2006", SAMPLE DURATION = "24 HOUR", "24 HOURS", "24-HR
BLK AVG" or "24-HR BLK" , UNITS OF MEASURE = "Micrograms/cubic meter (25
C)", 1ST MAX VALUE > 155

•	S02 1 -hour standard: Either PARAMETER CODE = 42401, POLLUTANT
STANDARD = "S02 1-hour 2010", SAMPLE DURATION = "1 HOUR", UNITS OF
MEASURE = "Parts per million", 1ST MAX VALUE > 0.0755 Or PARAMETER
CODE = 42401, POLLUTANT STANDARD = "S02 1-hour 2010", SAMPLE
DURATION = "1 HOUR", UNITS OF MEASURE = "Parts per billion", 1ST MAX
VALUE > 75.5

•	PM2.5 24-hour standard: PARAMETER CODE = 88101, POLLUTANT STANDARD =
"PM25 24-hour 2012", SAMPLE DURATION = "24 HOUR" or "24-HR BLK AVG",
UNITS OF MEASURE = "Micrograms/cubic meter (LC)", 1ST MAX VALUE > 35.5

The monitor exceeds the level of the standard if the corresponding set of conditions holds for the
given year. A county exceeds the level of the standard in a given year if any of the monitors in
that county exceed the level of the standard.

For the PM2.5 annual standard, summary data used for NAAQS attainment/non-attainment
designations for the years 2011 to 2016 were obtained directly from EPA in October and
November 2017 (data for the years 1999 to 2010 were previously obtained from EPA).V We
extracted the state and county FIPS codes for all years. These were used together with the
variable PM25wtdamviol, which has the value 1 if the county has a concentration above the level

America's Children and the Environment, Third Edition	Page 6

January 2018

-------
Environments and Contaminants: Criteria Air Pollutants

of the PM2.5 annual standard. In this case, a county exceeds the level of the standard if, at any
monitor in the county, the NAAQS data completeness requirements are met and the weighted
annual mean is above the target value of 12.05 |ig/m3. The weighted annual mean is the average
of the four quarterly means. This approach often gives different results from the values for PM2.5
annual standard in the Annual Summary files, which do not apply data completeness rules and
are based on the annual arithmetic mean of all the daily values.

For the lead standard, summary data with the annual maximum rolling three-month average lead
concentrations for each monitor from 1999 to 2006 were obtained directly from EPA on October
14, 2009.V1 These values were calculated using the completeness criteria and averaging
calculations detailed in the Federal Register, Vol. 73, No. 219. Thus, the maximum rolling three-
month average is calculated by finding all 12 three-month average lead concentrations that begin
in that calendar year, and finding the maximum of those 12 averages. For the year 2007, rolling
three-month average lead concentrations were obtained from the Excel file
Pb_DesignValues_20072009_Final.xls at the link for "Lead Detailed Information" for the end
year 2009 from the URL: http://epa.gov/airtrends/values previous.html. For the year 2008,
rolling three-month average lead concentrations were obtained from the Excel file for the end
year 2010 Pb_DesignValues_20082010_FinalRevised.xlsx at the same URL. For the year 2009,
rolling three-month average lead concentrations were obtained from the Excel file for the end
year 2011 Pb_DesignValues_2009201 1 Final.xlsx at the same URL. For the year 2010, rolling
three-month average lead concentrations were obtained from the Excel file for the end year 2012
Pb_DesignValues_20102012_Final_07_22_13.xlsx at the same URL. For the year 2011, annual
maximum rolling three-month average lead concentrations were obtained from the Excel file for
the end year 2013 Pb_DesignValues_20112013_Final_03_13_14.xlsx at the same URL. For the
year 2012, annual maximum rolling three-month average lead concentrations were obtained from
the Excel file for the end year 2014 Pb_DesignValues_20122014_FINAL_08_03_15.xlsx at the
URL: http://epa.gov/airtrends/values.html. For the year 2013, annual maximum rolling three-
month average lead concentrations were obtained from the Excel file for the end year 2015
Pb_DesignValues_20132015_FINAL_08_29_16.xlsx at the URL:

http://epa.gov/airtrends/values.html. For the years 2014 to 2016, annual maximum rolling three-
month average lead concentrations were obtained from the Excel file for the end year 2016
Pb_DesignValues_20142016_final_29junel7.xlsx at the URL: http://www.epa.gov/air-
trends/air-quality-design-values.

The annual maximum rolling three-month average lead concentrations for 2007 to 2010 were
calculated from the monthly values. The monitor has a concentration above the level of the
standard for the given year if the annual maximum rolling three-month average lead
concentration is at least 0.155 |ig/m3. A county is considered to have a concentration above the
level of the standard in a given year if any of the monitors in that county exceed the level of the
standard.

V1 Mark Schmidt, EPA OAQPS, 919-541-2416. Schmidt.Mark@epamail.epa.gov. Current contact: Halil Cakir. EPA

America's Children and the Environment, Third Edition	Page 7

January 2018

-------
Environments and Contaminants: Criteria Air Pollutants

For indicator El, we also evaluated the counties exceeding Any Standard; i.e., counties with
pollutant concentrations above the levels of one or more of the eight NAAQS standards used for
indicator El.

For the indicator E2, we used the same Annual Summary files to obtain the numbers of days that
the ozone or PM2.5 standards were exceeded. For ozone we used the observations for
PARAMETER CODE = 44201, POLLUTANT STANDARD = "Ozone 8-hour 2015", SAMPLE
DURATION = "8-HR RUN AVG BEGIN HOUR" or "8-HR RUN", UNITS OF MEASURE =
"Parts per million" and the variable PRIMARY VIOLATION COUNT, giving the number of
days where the maximum 8-hour average concentration exceeded the NAAQS level. For PM2.5
we used the observations for PARAMETER CODE = 88101, POLLUTANT STANDARD =
"PM25 24-hour 2012", SAMPLE DURATION = "24 HOUR" or "24-HR BLK AVG" , UNITS
OF MEASURE = "Micrograms/cubic meter (LC)" and the variable PRIMARY VIOLATION
COUNT, giving the number of days where the 24-hour average concentration exceeded the
NAAQS level. For PM2.5 we only used the data from the lowest numbered POC at each
monitoring site. The number of exceedances for a county is calculated as the maximum number
of days with concentrations above the levels of the standards across all monitors in that county.

Census Data

For the trend analyses we obtained children's populations by county for each year from 1999-
2016.

For 1999, the source was U.S. Census Bureau files, Estimates of the Population of Counties by
Age and Sex: 1990-1999, August 30, 2000. The file headers were "(CO-99-9) Population
Estimates for Counties by Age and Sex: Annual Time Series July 1, 1990 to July 1, 1999."
http://www.census.gov/popest/data/counties/asrh/1990s/CQ-99-09.html. These files give county
populations by age and sex for 1990 to 1999. We summed these populations by year and county
across all ages 0 to 17 and both sexes.

For 2000-2009, we used the bridged-race intercensal population files obtained from the CDC
website:

National Center for Health Statistics. Intercensal estimates of the resident population of the
United States for July 1, 2000-July 1, 2009, by year, county, single-year of age (0, 1, 2, .., 85
years and over), bridged race, Hispanic origin, and sex. Prepared under a collaborative
arrangement with the U.S. Census Bureau. Available from:

http://www.cdc.gov/nchs/nvss/bridged race.htm as of October 26, 2012, following release by
the U.S. Census Bureau of the revised unbridged intercensal estimates by 5-year age group
on October 9, 2012.

America's Children and the Environment, Third Edition
January 2018

Page 8

-------
Environments and Contaminants: Criteria Air Pollutants

Indian or Alaska Native, Asian or Pacific Islander), Hispanic origin (not Hispanic or Latino,
Hispanic or Latino), and sex. There is one SAS file for the years 2000-2004 and another SAS file
for the years 2005-2009. Files are available in SAS dataset and text formats; we used the SAS
dataset format for these analyses.

We extracted the following variables: state, county, age, racesex, hisp, and pop2000 to pop2009.
The racesex variable is a single coded value for each combination of race and sex, e.g., racesex =
1 denotes White males. The value popXXXX gives the population as of July 1 of the calendar
year XXXX for a given state, county, age, racesex combination, and ethnicity. The county
children's populations for each year 2000-2009 were obtained by summing the variable
popXXXX over all ages <= 17, all values of "racesex," and all values of "hisp." The county
children's populations for each year 2000-2009 and each race/ethnicity group were obtained by
summing the variable popXXXX over all ages <=17 for the applicable values of "racesex" and
"hisp."

For 2010 to 2016, we used the bridged race Vintage 2016 postcensal population file:

National Center for Health Statistics. Vintage 2016 postcensal estimates of the resident
population of the United States (April 1, 2010, July 1, 2010-July 1, 2016), by year, county,
single-year of age (0, 1,2,.., 85 years and over), bridged race, Hispanic origin, and sex.
Prepared under a collaborative arrangement with the U.S. Census Bureau. Available from:
http://www.cdc.gov/nchs/nvss/bridged_race.htm as of June 26, 2017, following release by the
U.S. Census Bureau of the unbridged Vintage 2016 postcensal estimates by 5-year age group
on June 22, 2017.

The bridged race 2010 to 2016 population files contain estimates of the resident population of
the United States as of July 1, 2010, July 1, 2011, July 1, 2012, July 1, 2012, July 1, 2013, July 1,
2014, July 1, 2015, and July 1, 2016 by county, single-year of age (0, 1, 2,..., 85 years and over),
bridged-race category (White, Black or African American, American Indian or Alaska Native,
Asian or Pacific Islander), Hispanic origin (not Hispanic or Latino, Hispanic or Latino), and sex.
Files are available in SAS dataset and text formats; we used the SAS dataset format for these
analyses.

We extracted the following variables: state, county, age, racesex, hisp, pop2010, pop2011,
pop2012, pop2013, pop2014, pop2015, and pop2016. The racesex variable is a single coded
value for each combination of race and sex, e.g., racesex = 1 denotes White males. The value
popXXXX gives the population as of July 1, XXXX for a given state, county, age, racesex
combination, and ethnicity. The county children's populations for the years 2010 to 2016 were
obtained by summing the variables pop2010, pop2011, pop2012, pop2013, pop2014, pop2015,
and pop2016 over all ages <= 17, all values of "racesex," and all values of "hisp." The county
children's populations for each year 2010-2016 and each race/ethnicity group were obtained by
summing the variable popXXXX over all ages <=17 for the applicable values of "racesex" and
"hisp."

America's Children and the Environment, Third Edition
January 2018

Page 9

-------
Environments and Contaminants: Criteria Air Pollutants

Calculation of Indicator

Indicator El is calculated as follows. Define, for each NAAQS,

Pop (county C, year Y) = population of children in county C, year Y

Exceed (county C, year Y) = 1, if county C has a pollutant concentration greater than the level of
the NAAQS in year Y

= 0, otherwise

These values are calculated as described above. Note that Exceed = 0 if the county has no air
quality monitors or no air quality monitoring data in the given year. Counties outside the 50
states or Washington, DC were excluded.

The total number of children living in counties in which concentrations were above the levels of
air quality standards is the sum of Pop over all counties where Exceed equals 1; i.e.,

Children Affected (year Y) = E [Pop (county C, year Y) x Exceed (county C, year Y)]

The total number of children living in the United States is the sum of Pop over all counties. Thus,

Children (year Y) = E Pop (county C, year Y)

The percentage of children living in counties in which pollutant concentrations were above the
levels of air quality standards equals the total number of children living in counties in which
concentrations were above the levels of air quality standards divided by the total number of
children living in the United States and multiplied by 100.

Percentage Children Affected = [Children Affected (year Y) / Children (year Y)] x 100%

Indicator E2 is calculated as follows. Define, for each NAAQS and each exceedance range a to b
(e.g., 11 to 25, or 26 to infinity),

Pop (county C, year Y) = population of children in county C, year Y.

ExceedRange (county C, year Y) = 1, if county C has K exceedances of the

NAAQS in year Y and a < K < b,
= 0, otherwise

These values are calculated as described above. Note that ExceedRange = 0 if the county has no
air quality monitors or no air quality monitoring data in the given year. Counties outside the 50
states or Washington, DC were excluded.

America's Children and the Environment, Third Edition
January 2018

Page 10

-------
Environments and Contaminants: Criteria Air Pollutants

The total number of children living in counties in which there were between a and b exceedances
of the levels of air quality standards is the sum of Pop over all counties where ExceedRange
equals 1; i.e.,

Children Affected (year Y) = E [Pop (county C, year Y) x ExceedRange (county C, year Y)]
The total number of children living in the United States is the sum of Pop over all counties. Thus,
Children (year Y) = E Pop (county C, year Y)

The percentage of children living in counties in which there were between a and b exceedances
of the levels of air quality standards equals the total number of children living in counties in
which there were between a and b exceedances of the levels of air quality standards divided by
the total number of children living in the United States and multiplied by 100.

Percentage Children Affected = [Children Affected (year Y) / Children (year Y)] x 100%

Race/Income

The supplementary tables Ela and Elb give the percentages of children living in counties in
which pollutant concentrations were above the levels of air quality standards, stratified by
race/ethnicity (Table Ela) or family income (Table Elb), for the year 2016 only. The calculation
of Tables Ela and Elb is exactly the same as for the trend analysis, except that the county
children's populations are replaced by estimated county children's populations for the given
race/ethnicity or income groups in 2016.

For the supplementary table Ela, children's populations stratified by race/ethnicity for 2016
were estimated for each county by using the bridged-race Vintage 2016 post-censal population
file for July 1, 2016 at the url http://www.cdc.gov/nchs/nvss/bridged race.htm to obtain
race/ethnicity distributions. The race/ethnicity groups were summed as follows:

•	White non-Hispanic: racesex = 1 and 2, hisp = 1, age <=17

•	Black non-Hispanic: racesex = 3 and 4, hisp = 1, age <=17

•	AIAN non-Hispanic: racesex = 5 and 6, hisp = 1, age <=17

•	API non-Hispanic: racesex = 7 and 8, hisp = 1, age <=17

•	Hispanic: racesex = 1, 2, ... 8, hisp = 2, age <=17

•	All: racesex = 1, 2, ... 8, hisp = 1 and 2, age <= 17

For the supplementary table Elb, children's populations stratified by income for 2016 were
obtained using data from the Census Bureau's Small Area Poverty and Income Estimates
(SAIPE) program combined with the populations from the bridged-race Vintage 2016 post-
censal population file for July 1, 2016. The SAIPE population estimates were developed by the

America's Children and the Environment, Third Edition
January 2018

Page 11

-------
Environments and Contaminants: Criteria Air Pollutants

Census Bureau using the American Community Survey (ACS) and Census 2010 data, based on
statistical models. The SAIPE data were obtained from the url

https://www. census, gov/data/datasets/2016/demo/saipe/2016-state-and-countv.html

as the Excel file estl6ALL.xls (downloaded in December, 2017). The Excel file includes the
state and county FIPS codes and the poverty percentage under age 18. For each county, the total
children's population was calculated by summing populations across all children ages 0 to 17
years. The children's population below the poverty level was obtained by multiplying the total
children's population by the poverty percentage under age 18. The children's population at or
above the poverty level was obtained by multiplying the total children's population by 100 minus
the poverty percentage under age 18:

Population below poverty = Total population x poverty percentage under age 18/100,

Population at or above poverty =

Total population x (100 - poverty percentage under age 18)/100.

Questions and Comments

Questions regarding these methods, and suggestions to improve the description of the methods,
are welcome. Please use the "Contact Us" link at the bottom of any page in the America's
Children and the Environment website.

Statistical Comparisons

Statistical analyses of the percentages of children ages 0 to 17 years living in counties with
concentrations above the levels of air quality standards were used to determine whether the
trends in the percentages for different years were statistically significant. Using a logistic
regression model, the logarithm of the odds that a child lives in a county with concentrations
above the levels of air quality standards is regressed against the calendar year. To avoid potential
statistical issues associated with very low or very high percentages, each percentage was first
rescaled to be between 5% and 95% before computing the odds. The error terms for this logistic
regression were assumed to be approximately independent and normally distributed. For each
criteria air pollutant, the slope of the regression line for the logarithm of the odds was computed,
together with a 95% confidence interval and its p-value. The slope estimates the annual change
in the logarithm of the odds. A p-value at or below 0.05 implies that the trend is statistically
significant at the 5% significance level. No adjustment is made for multiple comparisons.

The results are presented in Table 2. For more details on these statistical analyses, see the
memorandum by Cohen (2010).vu

vu Statistical methods for testing for trends and year-to-year changes to air quality measures. Memorandum from
Jonathan Cohen, ICF, to Dan Axelrad, EPA, November 2010.

America's Children and the Environment, Third Edition Page 12

January 2018

-------
Environments and Contaminants: Criteria Air Pollutants

Table 2. Logistic regression trend test for the proportion of children ages 0 to 17 years
living in counties in which pollutant concentrations were above the levels of air quality
standards in years 1999 to 2016.

Pollutant

N

Trend (annual

95%

95% Confidence

P-value for





change in log

Confidence

Interval for Trend:

Trend





odds)

Interval for
Trend:
Lower
Bound

Upper Bound



Ozone
(eight-hour)

18

-0.027

-0.040

-0.014

<0.001

PMio (24-
hour)

18

-0.019

-0.036

-0.002

0.031

PM2s (24-
hour)

18

-0.106

-0.129

-0.082

<0.001

PM2s
(annual)

18

-0.149

-0.180

-0.119

<0.001

Carbon
monoxide

18

-0.034

-0.051

-0.018

<0.001

Lead

18

-0.003

-0.029

0.023

0.809

Sulfur











dioxide
(one-hour)

18

-0.106

-0.115

-0.097

<0.001

Nitrogen

dioxide

(one-hour)

18

-0.081

-0.099

-0.063

<0.001

Any

standard

18

-0.045

-0.058

-0.032

<0.001

America's Children and the Environment, Third Edition
January 2018

Page 13

-------
Environments and Contaminants: Criteria Air Pollutants

Methods

Indicator

E3. Percentage of days with good, moderate, or unhealthy air quality for children ages 0 to 17
years, 1999-2015.

Summary

EPA's Air Quality Index (AQI) classifies air quality as good, moderate, or unhealthy on a daily
basis, for each county in the United States. The AQI is based on monitoring data for six criteria
air pollutants. For each county, year, and day, the air quality index category for that day was
obtained from EPA OAQPS. For each county and year, the annual numbers of days categorized
as good, moderate, or unhealthy were counted. Indicator E3 is the national weighted average of
these annual numbers, where the weights are the numbers of children ages 0 to 17 years living in
each county, obtained from the Census Bureau. This is mathematically equivalent to the
percentage of children's days in each category. The supplementary tables for indicator E3
provide the percentages of days for children ages 0 to 17 years with good, moderate, or
unhealthy air quality, stratified by race/ethnicity (Table E3a) or income (Table E3b), for the year
2015.

Overview of Data Files

The following files are needed to calculate this indicator:

• Daily monitoring and AQI data. This file contains the state and county FIPS codes, site
id, POC, date, parameter code, daily maximum, event type, and the corresponding AQI
for each parameter code.

• Census data. This file contains the state and county FIPS codes, year, and children's
population. For 1999, we obtained this information from the U.S. Census Bureau files:

Estimates of the Population of Counties by Age and Sex: 1990-1999, August 30,
2000. The file headers were "(CO-99-9) Population Estimates for Counties by Age
and Sex: Annual Time Series July 1, 1990 to July 1, 1999."
http://www.census.gov/popest/data/counties/asrh/1990s/CQ-99-09.html.

These files give county populations by age and sex for 1990 to 1999. We summed
these populations by year and county across all ages 0 to 17 years and both sexes.

For 2000-2009, we obtained this information from the bridged-race intercensal
population files:

National Center for Health Statistics. Intercensal estimates of the resident
population of the United States for July 1, 2000-July 1, 2009, by year, county, single-

America's Children and the Environment, Third Edition
January 2018

Page 14

-------
Environments and Contaminants: Criteria Air Pollutants

year of age (0, 1, 2,85 years and over), bridged race, Hispanic origin, and sex.
Prepared under a collaborative arrangement with the U.S. Census Bureau. Available
from: http://www.cdc.gov/nchs/nvss/bridged race.htm as of October 26, 2012,
following release by the U.S. Census Bureau of the revised unbridged intercensal
estimates by 5-year age group on October 9, 2012.

The children's populations by year and county were obtained by summing across the ages
0 to 17 years inclusive.

For 2010-2015, we obtained this information from the bridged race Vintage 2015
postcensal population file:

National Center for Health Statistics. Vintage 2015 postcensal estimates of the
resident population of the United States (April 1, 2010, July 1, 2010-July 1, 2015), by
year, county, single-year of age (0, 1, 2,.., 85 years and over), bridged race, Hispanic
origin, and sex. Prepared under a collaborative arrangement with the U.S. Census
Bureau. Available from: http://www.cdc.gov/nchs/nvss/bridged race.htm , as of June
24, 2015, following release by the U.S. Census Bureau of the unbridged Vintage
2015 postcensal estimates by 5-year age group on June 23, 2015.

The children's populations by year, county, and race/ethnicity were obtained by summing
across the ages 0 to 17 inclusive.

Air Quality Data

Air Quality Index (AQI) and daily maximum monitoring data from the EPA Air Quality System
(AQS) for the years 1999 to 2015 and for 8-hour ozone, 24-hour PMio, 24-hour PM2.5, 8-hour
CO, 1-hour SO2, and 1-hour NO2 were downloaded in March 2017 from the daily summary files
at the EPA Air Data website:

http s: //aq s. epa. gov/aq sweb/airdata/downl oad fil e s. html

AQI monitoring data for the years 1999-2015 and for 1-hour ozone and 24-hour SO2 were
obtained directly from OAQPS staff.V1U These daily summary files give the air quality index for
each monitor, day, and pollutant standard for which the AQI is defined. The Event Type variable
for a criteria air pollutant flags any days deemed to have exceptional events (e.g., forest fires).
For these analyses, any exceptional event days were included because we excluded data with the
Event Type "Excluded." For PM2.5, if a monitoring site had AQI values for multiple POCs on the
same date, then the AQI for the lowest POC was used. For the 8-hour ozone standard, all the
AQI values were recomputed using the daily maximum 8-hour average ozone values, as
described below. The AQI for a given county on a given day is obtained by determining the
maximum of all available AQI values for that day across all monitors in the county and across all
criteria air pollutants in the index. The following air quality index categories are assigned based

America's Children and the Environment, Third Edition Page 15

January 2018

-------
Environments and Contaminants: Criteria Air Pollutants

on the breakpoints shown in Table 3: Good, Moderate, and Unhealthy (includes Unhealthy for
Sensitive Groups, Unhealthy, Very Unhealthy, and Hazardous).

Table 3 shows the calculation of the air quality index for each day and criteria air pollutant,
based on the measured concentrations on that day. The overall AQI category for a given monitor
and day is defined as the highest AQI category among all the pollutants measured on that day, if
any. The overall AQI category for a given county and day is defined as the highest AQI category
among all the monitors in the county with air quality measurements on that day, if any.

Table 3. Breakpoints for the AQI

This Breakpoint...

...equals
this
AQI

...and this
category

03 (ppm)
8-hour
Code 44201

O3
(ppm)
1-hour3
Code
44201

PM10
(|ig/m3)
24-hour
Code
81102

PM25

(|ig/m3)
24-hour
Code
88101

CO
(ppm)
8-hour
Code
42101

S02
(ppb)
1-hour
Code
42401

N02

(ppb)

1-hour

Code

42602

AQI



0.000 - 0.054

~

0-54

0.0-12.0

0.0-
4.4

0-35

0-53

0-50

Good

0.055-0.070

~

55-
154

12.1-35.4

4.5-
9.4

36-75

54 - 100

51-
100

Moderate

0.071-0.085

0.125-
0.164

155 -
254

35.5 -55.4

9.5-
12.4

76 - 185

101 -

360

101 -
150

Unhealthy

for
Sensitive
Groups

0.086-0.105

0.165-
0.204

255-
354

55.5-
150.4

12.5-
15.4

186-
304°

361 -
649

151 -

200

Unhealthy

0.106-0.200

0.205 -
0.404

355 -
424

150.5-
250.4

15.5-
30.4

305 -
604°

650-
1249

201-
300

Very
Unhealthy

b

0.405 -
0.504

425-
504

250.5 -
350.4

30.5-
40.4

605 -
804°

1250-
1659

301 -
400

Hazardous

b

0.505 -
0.604

505-
604

350.5 -
500.4

40.5-
50.4

805 -
1004°

1650-
2049

401-
500

Hazardous

Adapted from "Technical Assistance Document for the Reporting of Daily Air Quality - the Air Quality Index
(AQI)," EPA-454/B-16-002, May 2016. Available at http://www.airnow.gov/index.cfm?action=pubs.index.

a Areas are required to report the AQI based on 8-hour ozone values. However, there are areas where an AQI based
on 1-hour ozone values would be more protective. In these cases the index for both the 8-hour and the 1-hour ozone
values may be calculated and the maximum AQI reported.

b 8-hour O3 values do not define higher AQI values (> 301). AQI values of 301 or higher are calculated with 1-hour
O3 concentrations.

0 1-hour SO2 values do not define higher AQI values (> 200). AQI values of 200 or higher are calculated with 24-
hour SO2 concentrations.

For the 8-hour ozone standard, many of the AQI values in the daily files downloaded in March
2017 were erroneously calculated using the previous ozone standard. The daily AQI values were

America's Children and the Environment, Third Edition
January 2018

Page 16

-------
Environments and Contaminants: Criteria Air Pollutants

recomputed from the daily maximum values using the breakpoints given in Table 3. For each
category in Table 3, let BPlo and BPm denote the lower and upper breakpoints for 8-hour ozone
given in the first column of Table 3, and let Ilo and Im denote the lower and upper AQI values in
the AQI column. Let C denote the daily maximum 8-hour average ozone concentration truncated
to 3 decimal places. For each C, we found the category so that BPm is greater than or equal to C
and BPlo is less than or equal to C. We used the following equation to compute the AQI for the
8-hour ozone standard:

AQ'=bp"~bp (c-bp^+'i-

Hi Lo

The AQI is rounded to the nearest integer.1X

For the calculations for indicator E3, the following overall AQI categories were combined into
the category "Unhealthy": Unhealthy for Sensitive Groups, Unhealthy, Very Unhealthy, and
Hazardous.

For each county and year, we counted the total number of days in the good, moderate, and
unhealthy categories and also the number of days without data (by subtraction from the number
of days in each year). Counties not in the 50 states or Washington, DC were excluded. For
county/year combinations not in the AQI data, we assumed zero days in the good, moderate, and
unhealthy categories and 365 (366 for leap years) days without data.

Census Data

For the trend analyses we obtained children's populations by county for each year from 1999 to
2015.

For 1999, the source was U.S. Census Bureau files:

Estimates of the Population of Counties by Age and Sex: 1990-1999, August 30, 2000.
The file headers are "(CO-99-9) Population Estimates for Counties by Age and Sex:
Annual Time Series July 1, 1990 to July 1, 1999."
http://www.census.gov/popest/data/counties/asrh/1990s/CQ-99-09.html.

These files give county populations by age and sex for 1990 to 1999. We summed these
populations by year and county across all ages 0 to 17 years and both sexes.

For 2000-2009, we used the bridged-race intercensal population files obtained from the CDC
website:

lxSee Equation 1 in "Technical Assistance Document for the Reporting of Daily Air Quality - the Air Quality Index

America's Children and the Environment, Third Edition Page 17

January 2018

-------
Environments and Contaminants: Criteria Air Pollutants

National Center for Health Statistics. Intercensal estimates of the resident population of the
United States for July 1, 2000-July 1, 2009, by year, county, single-year of age (0, 1, 2, 85
years and over), bridged race, Hispanic origin, and sex. Prepared under a collaborative
arrangement with the U.S. Census Bureau. Available from:

The bridged-race intercensal population files contain estimates of the resident population of the
United States as of July 1, 2000; July 1, 2001; July 1, 2002; July 1, 2003; July 1, 2004; July 1,
2005; July 1, 2006; July 1, 2007; July 1, 2008; and July 1, 2009 by county, single-year of age (0,
1, 2,..., 85 years and over), bridged-race category (White, Black or African American, American
Indian or Alaska Native, Asian or Pacific Islander), Hispanic origin (not Hispanic or Latino,
Hispanic or Latino), and sex. There is one SAS file for the years 2000-2004 and another SAS file
for the years 2005-2009. Files are available in SAS dataset and text formats; we used the SAS
dataset format for these analyses.

For 2010-2015, we obtained this information from the bridged race Vintage 2015 postcensal
population file:

National Center for Health Statistics. Vintage 2015 postcensal estimates of the resident
population of the United States (April 1, 2010, July 1, 2010-July 1, 2015), by year, county,
single-year of age (0, 1,2,.., 85 years and over), bridged race, Hispanic origin, and sex.
Prepared under a collaborative arrangement with the U.S. Census Bureau. Available from:
http://www.cdc.gov/nchs/nvss/bridged race.htm , as of June 24, 2015, following release by
the U.S. Census Bureau of the unbridged Vintage 2015 postcensal estimates by 5-year age
group on June 23, 2015.

The bridged race 2010 to 2015 population files contain estimates of the resident population of
the United States as of July 1, 2010, July 1, 2011, July 1, 2012, July 1, 2013, July 1, 2014, and
July 1, 2015 by county, single-year of age (0, 1, 2,..., 85 years and over), bridged-race category
(White, Black or African American, American Indian or Alaska Native, Asian or Pacific
Islander), Hispanic origin (not Hispanic or Latino, Hispanic or Latino), and sex. Files are
available in SAS dataset and text formats; we used the SAS dataset format for these analyses.

We extracted the following variables: state, county, age, racesex, hisp, pop2010, pop2011,
pop2012, pop2013, pop2014, and pop2015. The racesex variable is a single coded value for each
combination of race and sex, e.g., racesex = 1 denotes White males. The value popXXXX gives

America's Children and the Environment, Third Edition
January 2018

Page 18

-------
Environments and Contaminants: Criteria Air Pollutants

the population as of July 1, XXXX for a given state, county, age, racesex combination, and
ethnicity. The county children's populations for the years 2010, 2011, 2012. 2013, 2014, and
2015 were obtained by summing the variables pop2010, pop2011, pop2012, pop2013, pop2014,
and pop2015 over all ages <= 17, all values of "racesex," and all values of "hisp."

Calculation of Indicator

Indicator E3 is calculated as follows. The percentage of days is calculated directly for three
categories: Good, Moderate, or Unhealthy. The calculation is the same for each category; the
following example is for the Good category. Define

Days (category G, county C, year Y) = number of days in AQI category G for county C,
year Y

Pop (county C, year Y) = population of children in county C, year Y
These values are calculated as described above.

1.	The county number of children's days in category G is calculated as the number of G days in
county C times the number of children in county C:

County children's G days (G, C, Y) = Days (G, C, Y) x Pop (C, Y)

2.	The national number of children's days in category G is calculated by summing the county
numbers over the counties:

National children's G days (G, Y) = E County children's G days in G (G, C, Y)

where this sum is taken across all counties in all 50 states plus Washington, DC.

3.	The county number of children's days (in all categories) is calculated as the number of days in
year Y (365, or 366 for leap years) times the county number of children:

County children's days (C, Y) = 365 (or 366) x Pop (C, Y)

4.	The national number of children's days is calculated by summing the county numbers over the
counties:

National children's days (Y) = E County children's days (C, Y)
where this sum is taken across all counties in all 50 states plus Washington, DC.

America's Children and the Environment, Third Edition
January 2018

Page 19

-------
Environments and Contaminants: Criteria Air Pollutants

5. The percentage of children's days in category G is calculated by dividing the national
children's G days by the national children's days and multiplying by 100 (to make it a
percentage):

E3 = Percentage of children's days in category G

= [National children's G days (G, Y) / National children's days (Y)] x 100%

The numbers of unmonitored days are by definition the total number of good, moderate, and
unhealthy days subtracted from the number of days in the year. To calculate the percentage of
children's days that are unmonitored you may either use the same method applied to the category
of unmonitored days, or you may subtract the total of the percentages for good, moderate, and
unhealthy children's days from 100. Both approaches will give exactly the same answer.

Race/Income

The supplementary tables for Indicator E3 give the percentages of children's days with good,
moderate, or unhealthy air quality stratified by race/ethnicity (Table E3a) or family income
(Table E3b), for the year 2015 only. The calculation of Tables E3a and E3b is exactly the same
as for the trend analysis except that the county children's populations are replaced by estimated
county children's populations for the given race/ethnicity or income groups in 2015.

For the supplementary table E3a, children's populations stratified by race/ethnicity for 2015
were estimated for each county by using the bridged-race Vintage 2015 postcensal population
file for July 1. 2015 at the URL http://www.cdc.gov/nchs/nvss/bridged race.htm to obtain
race/ethnicity distributions. The race/ethnicity groups were summed as follows:

•	White non-Hispanic: racesex = 1 and 2, hisp = 1, age <=17

•	Black non-Hispanic: racesex = 3 and 4, hisp = 1, age <=17

•	AIAN non-Hispanic: racesex = 5 and 6, hisp = 1, age <=17

•	API non-Hispanic: racesex = 7 and 8, hisp = 1, age <=17

•	Hispanic: racesex = 1, 2, ... 8, hisp = 2, age <=17

•	All: racesex = 1, 2, ... 8, hisp = 1 and 2, age <= 17

For the supplementary table E3b, children's populations stratified by income for 2015 were
obtained using data from the Census Bureau's Small Area Poverty and Income Estimates
(SAIPE) program combined with the populations from the bridged-race Vintage 2015 postcensal
population file for July 1, 2015. The SAIPE population estimates were developed by the Census
Bureau using the American Community Survey (ACS) and Census 2010 data, based on statistical
models. The SAIPE data were obtained from the URL

http ://www. census. gov/did/www/ saipe/data/ statecountv/data/2015 .html

America's Children and the Environment, Third Edition
January 2018

Page 20

-------
Environments and Contaminants: Criteria Air Pollutants

as the Excel file estl5ALL.xls (downloaded in December, 2016). The Excel file includes the
state and county FIPS codes and the poverty percentage under age 18. For each county, the total
children's population was calculated by summing populations across all children ages 0 to 17
years. The children's population below the poverty level was obtained by multiplying the total
children's population by the poverty percentage under age 18. The children's population at or
above the poverty level was obtained by multiplying the total children's population by 100 minus
the poverty percentage under age 18:

Population below poverty = Total population x poverty percentage under age 18/100,

Population at or above poverty =

Total population x (100 - poverty percentage under age 18)/100.

Questions and Comments

Statistical Comparisons

Statistical analyses of the percentages of days for children ages 0 to 17 years with good,
moderate, or unhealthy air quality were used to determine whether the trends in the percentages
for different years were statistically significant. Using a logistic regression model, the logarithm
of the odds that a children's day has good, moderate, or unhealthy air quality is regressed against
the calendar year. To avoid potential statistical issues associated with very low or very high
percentages, each percentage was first rescaled to be between 5% and 95% before computing the
odds. The error terms for this logistic regression were assumed to be approximately independent
and normally distributed. For each air quality index category, the slope of the regression line for
the logarithm of the odds was computed, together with a 95% confidence interval and its p-value.
The slope estimates the annual change in the logarithm of the odds. A p-value at or below 0.05
implies that the trend is statistically significant at the 5% significance level. No adjustment is
made for multiple comparisons.

The results are presented in Table 4. For more details on these statistical analyses, see the
memorandum by Cohen (2010).x

x Statistical methods for testing for trends and year-to-year changes to air quality measures. Memorandum from
Jonathan Cohen, ICF, to Dan Axelrad, EPA, November 2010.

America's Children and the Environment, Third Edition Page 21

January 2018

-------
Environments and Contaminants: Criteria Air Pollutants

Table 4. Logistic regression trend test for the proportion of days for children ages 0 to 17
years with good, moderate, or unhealthy air quality in years 1999 to 2015.





Trend (annual
change in log
odds)

95% Confidence

95% Confidence



Pollution Level

N

Interval for
Trend: Lower

Interval for
Trend: Upper

P-value for
Trend





Bound

Bound



Good

17

0.0388

0.0339

0.0437

<0.001

Moderate

17

-0.0221

-0.0269

-0.0173

<0.001

Unhealthy

17

-0.0444

-0.0524

-0.0363

<0.001

No Monitoring
Data

17

-0.0051

-0.0078

-0.0023

0.001

America's Children and the Environment, Third Edition
January 2018

Page 22

-------