Download:
pdf |
pdfDOT HS 811 111
April 2009
NCSA Technical Report:
The 2006 National Survey of the
Use of Booster Seats —
Methodology Report
This document is available to the public from the National Technical Information Service, Springfield, Virginia 22161
DISCLAIMER
This publication is distributed by the U.S. Department of Transportation, National
Highway Traffic Safety Administration, in the interest of information exchange. The
opinions, findings, and conclusions expressed in this publication are those of the authors
and not necessarily those of the Department of Transportation or the National Highway
Traffic Safety Administration. The United States Government assumes no liability for its
contents or use thereof. If trade names, manufacturers’ names, or specific products are
mentioned, it is because they are considered essential to the object of the publication and
should not be construed as an endorsement. The United States Government does not
endorse products or manufacturers.
This document is available for free download at www.nhtsa.gov
Technical Report Documentation Page
1. Report No.
2. Government Accession No.
3. Recipient's Catalog No.
DOT HS 811 111
4. Title and Subtitle
5. Report Date
The 2006 National Survey of the Use of Booster Seats – Methodology Report
April 2009
6. Performing Organization Code
NVS-421
7. Author(s)
8. Performing Organization Report No.
Glassbrenner, Donna, Ph.D.
9. Performing Organization Name and Address
10. Work Unit No. (TRAIS)
Mathematical Analysis Division, National Center for Statistics and Analysis
National Highway Traffic Safety Administration
U.S. Department of Transportation, NVS-421
1200 New Jersey Avenue SE.
Washington, DC 20590
11. Contract or Grant No.
DTNH22-05-D-01002
12. Sponsoring Agency Name and Address
13. Type of Report and Period Covered
Mathematical Analysis Division, National Center for Statistics and Analysis
National Highway Traffic Safety Administration
U.S. Department of Transportation, NVS-421
1200 New Jersey Avenue SE.
Washington, DC 20590
NHTSA Technical Report
14. Sponsoring Agency Code
15. Supplementary Notes
The 2006 National Survey of the Use of Booster Seats was conducted by WESTAT, Inc., under Federal contract with NHTSA.
A number of portions of this report are paraphrasings of WESTAT’s training material and final report to NHTSA from the
2006 survey.
We thank Dr. Eun Young Noh of URC Enterprises, Inc., for her assistance in preparing this report.
Abstract
The purpose of this report is to document the survey design used for the initial data collection of the National
Survey of the Use of Booster Seats (NSUBS). The initial data collection occurred in 2006. Although this report is
being published after the second data collection (which occurred in 2007) and as NHTSA is preparing for the third
data collection (in 2008), this report serves as important documentation as the design used for subsequent NSUBS
data collections were based on the design documented in this report, incorporating relatively minor changes in
methodology. NHTSA expects to publish annual methodology reports that document subsequent changes in
methodology.
17. Key Words
18. Distribution Statement
Child restraints, car seats, booster seats, survey
methodology
This document is available for free download at
www.nhtsa.gov. Printed copies can also be purchased by
contacting the National Technical Information Service,
Springfield, VA 22161, or visiting www.ntis.gov.
19. Security Classif. (of this report)
20. Security Classif. (of this page)
21. No. of Pages
Unclassified
Unclassified
73
Form DOT F 1700.7 (8-72)
Reproduction of completed page authorized
i
22. Price
TABLE of CONTENTS
1.
Summary ....................................................................................................................... 1
2.
The Circumstances That Gave Rise to the NSUBS ...................................................... 2
2.1
Booster Age Children – An Area of Particular Concern.............................................. 2
2.2
The Transportation Recall Enhancement, Accountability, and Documentation Act of
2000 (TREAD) .............................................................................................................. 2
2.3
A Data Need.................................................................................................................. 2
3.
The Prior State of Knowledge - A History of Booster Seat Use Estimates.................. 4
3.1
What’s Difficult About Getting a Reliable Estimate of Booster Seat Use .................... 4
3.2
Estimates of Booster Seat Use Prior to NSUBS............................................................ 4
3.3
Considering an Optimal Survey.................................................................................... 5
4.
Sample Design .............................................................................................................. 6
4.1
Selection of Primary Sampling Units............................................................................ 6
4.1.1 PSU Sampling Frame ......................................................................................... 6
4.1.2 Selection of PSUs ............................................................................................... 6
4.2
Selection of Sites Within PSUs...................................................................................... 7
4.2.1 Site Sampling Frame........................................................................................... 7
4.2.2 Selection of the Probability Sample of Sites .................................................... 10
4.2.3 Post-Selection Refinements to the Probability Sample: Obtaining the
“Refined Sample” ...................................................................................................... 11
4.2.4 Post-Selection Process to Identify Duplicates in the Second Stage Frame ...... 12
4.3
Site Selection Probabilities ......................................................................................... 13
4.4
Sample Size Determination ......................................................................................... 16
5.
Data Collection Protocols ........................................................................................... 19
5.1
Obtaining Site Cooperation ........................................................................................ 19
5.1.1 How Cooperation Was Obtained ...................................................................... 19
5.1.2 Site Participation Rates for the 2006 Survey.................................................... 19
5.2
The Data Collection Schedule .................................................................................... 20
5.3
Number, Gender, and Positioning of Data Collectors ............................................... 21
5.3.1 Number and Gender of Data Collectors ........................................................... 21
5.3.2 Positioning Data Collectors at the Site to Best Collect Data............................ 21
5.4
The Definition of Restraint Use Used by the Survey................................................... 23
NHTSA’s National Center for Statistics and Analysis, 1200 New Jersey Avenue SE., Washington, DC
20590
ii
5.5
The Categories of Race and Ethnicity Used by the Survey......................................... 24
5.6
The Survey Variables .................................................................................................. 24
5.7
The Wording of Interview Questions .......................................................................... 27
5.8
Other Assorted Data Collection Topics and Definitions ............................................ 28
5.8.1 Items Provided to Data Collectors ................................................................ 28
5.8.2 The Types of Vehicles Surveyed .................................................................. 29
5.8.3 Keeping Track of Nonparticipating Vehicles ............................................... 29
5.8.4 Miscellaneous ............................................................................................... 29
5.9
Data Collection Protocols .......................................................................................... 30
6.
Quality Control Procedures......................................................................................... 34
6.1
Pilot Testing of Data Collection Protocols................................................................. 34
6.2
Recruitment of Field Staff ........................................................................................... 34
6.3
Training....................................................................................................................... 34
6.4
Pre-Collection Test of Data Collectors ...................................................................... 35
6.5
Contact Information for Questions ............................................................................. 35
6.6
Unannounced Site Visits ............................................................................................. 35
7.
Data Entry, Editing, and Imputation ........................................................................... 36
7.1
Data Entry................................................................................................................... 36
7.2
Editing......................................................................................................................... 37
7.3
Imputation ................................................................................................................... 37
7.3.1 Variables Imputed as Special Cases ............................................................. 37
7.3.2 Variables Imputed by Logical Imputation .................................................... 37
7.3.3 Variables Not Imputed.................................................................................. 38
7.3.4 Variables Imputed by Hot-deck Imputation ................................................. 38
8.
Estimation ................................................................................................................... 39
8.1
Estimator Design ........................................................................................................ 39
8.2
Adjustment for Variation in Duration of Data Collection .......................................... 41
8.3
Nonresponse Adjustment Factors ............................................................................... 41
8.4
Weight Trimming ........................................................................................................ 43
8.5
The Estimation Formula ............................................................................................. 44
8.6
Estimates Computed.................................................................................................... 45
8.7
Definitions of Categories Used in Estimates .............................................................. 45
8.8
A Note on the Race/Ethnicity Estimates...................................................................... 46
NHTSA’s National Center for Statistics and Analysis, 1200 New Jersey Avenue SE., Washington, DC
20590
iii
9.
Variance Estimation.................................................................................................... 47
10.
Rules for Suppressing Estimates in Publications........................................................ 50
11.
Glossary of Terms....................................................................................................... 51
12.
Glossary of Notation ................................................................................................... 54
13.
References................................................................................................................... 58
14.
Appendix..................................................................................................................... 60
14.1
Data Collection Forms ............................................................................................... 60
14.2
Letter of Authorization................................................................................................ 61
14.3
Incentives .................................................................................................................... 63
14.4
Frequently Asked Questions ....................................................................................... 63
14.5
Card for Race/Ethnicity Questions ............................................................................. 65
NHTSA’s National Center for Statistics and Analysis, 1200 New Jersey Avenue SE., Washington, DC
20590
iv
TABLE of FIGURES
A Collection of Booster Seat Use Estimates................................................................................... 5
The NSUBS Probability Sample and Refined Sample ................................................................. 12
Desired Margins of Error .............................................................................................................. 16
Achieved Margins of Error ........................................................................................................... 17
Business Recruitment Results by Site Type ................................................................................. 20
Location of Data Collectors at the Sites ....................................................................................... 22
Definitions Used for the Survey Variable “Restraint Used” ........................................................ 23
Survey Variables Collected........................................................................................................... 24
Wording of Interview Questions................................................................................................... 27
Variables Imputed as Special Cases ............................................................................................. 37
Estimates Whose Variances Were Estimated Directly ................................................................. 47
Booster Seat Recording Form....................................................................................................... 60
Incentives ...................................................................................................................................... 62
NHTSA’s National Center for Statistics and Analysis, 1200 New Jersey Avenue SE., Washington, DC
20590
v
1.
Summary
In 2006, NHTSA conducted the first-ever nationwide survey of booster seat use in the United States
based on the observation of children in vehicles – the National Survey of the Use of Booster Seats, or
NSUBS. The survey presented challenges in developing an appropriate sample design, data collection
protocols, and statistical estimation.
The purpose of this publication is to present the choices made to address these challenges and fully
document the design of the survey used for the 2006 data collection. Although this report is being
published after the second data collection (which occurred in 2007) and as NHTSA is preparing for the
third data collection (in 2008), this report serves as important documentation as the design used for
subsequent NSUBS data collections were based on the design documented in this report, incorporating
relatively minor changes in methodology. NHTSA expects to publish annual methodology reports that
document the design used for any given data collection and identify any design-related changes made
since the prior data collection.
The portions of this report on sample design, editing, nonresponse adjustment, estimation, and variance
estimation are written for a statistical audience. A simpler description of the sample design that leaves
out many of the details for the statistical audience may be found in Glassbrenner and Ye, 2007, and its
annual updates on www.nhtsa.gov.
The NSUBS is conducted by the National Center for Statistics and Analysis (NCSA), an office of the
National Highway Traffic Safety Administration. The survey design, data collection, data editing,
nonresponse adjustment, and calculation of estimates and variances were conducted by WESTAT, Inc.,
under the direction of NCSA, via NHTSA contract number DTNH22-05-D-01002.
OMB approval was obtained for the collection of data for this survey. NHTSA obtained approval to
collect data for the 2006-2009 surveys under OMB clearance number 2127-0644. The notice of OMB
review can be found in the Federal Register, Volume 71, Number 30, page 7824, February 14, 2006.
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
1
2.
The Circumstances that Gave Rise to the NSUBS
2.1
Booster-Age Children – An Area of Particular Concern
Great strides have been made in recent years in protecting child passengers. Among infants and toddlers,
restraint use remains near 90 percent (98% for infants and 89% for children 1 to 3 years old in 2006), and
crash-related fatalities for 0 to 3-year-old occupants dropped by 13 percent in 2005, compared to 2004.
(Glassbrenner & Ye, February 2007; NHTSA, 2006)
Unfortunately, similar progress has not been achieved where older child passengers are concerned.
Booster seat use -- estimated at only 10 to 20 percent nationwide1 when the survey that is the subject of
this report was conducted -- remains unacceptably low. According to NHTSA’s Fatality Analysis
Reporting System (FARS) and National Automotive Sampling System General Estimates System (NASS
GES), in 2005 there were 346 fatalities among booster-age child passengers – children between the ages
of 4 and 7, inclusive -- as well as 49,000 injuries in this age group (NHTSA, 2006) Only 78 percent of
children 4 to 7 were restrained in 2005, a 5-percentage-point drop since 2002, according to NHTSA’s
National Occupant Protection Use Survey (NOPUS) (Glassbrenner & Ye, February 2007; Glassbrenner,
February 2005).
2.2 The Transportation Recall Enhancement, Accountability, and
Documentation Act of 2000
In 2000, Congress passed the Transportation Recall Enhancement, Accountability, and Documentation
(TREAD) Act of 2000. Section 14(i) of the act directs the Department of Transportation to reduce the
deaths and injuries among children in the 4- to 8-year-old age group that are caused by failure to use
booster seats by 25 percent. Conducting the National Survey of the Use of Booster Seats provides the
Department with invaluable information on who is and is not using booster seats, helping the Department
better direct its outreach programs to ensure that children are protected to the greatest degree possible
when they ride in motor vehicles. In particular, the information collected in this survey support the
Department of Transportation goal to improve safety in motor vehicle transportation.
Also in 2002, Congress enacted Public Law 107-318, known as Anton’s Law, which contains additional
provisions to improve the safety of child restraints in passenger motor vehicles, especially for older-child
passengers. [Public Law 107-318, Dec. 4, 2002]
In the TREAD Act and Anton’s Law, NHTSA was directed to conduct a range of initiatives, including
rulemaking, compliance testing, and consumer education programs, to enhance the safety of older child
passengers.
2.3
A Data Need
In order to adequately address the TREAD requirements, DOT needed data on who is and who is not
using booster seats in order to target outreach programs.
1
NHTSA estimated booster seat use to be in this range based on estimates from the Children’s Hospital of
Philadelphia (CHOP) in 2002 and NHTSA’s Motor Vehicle Occupant Safety Survey (MVOSS) in 2003. See
Partners for Child Passenger Safety, 2004, and Boyle et al., 2005, for more information on these estimates.
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
2
Previous estimates of booster seat use were not sufficiently reliable to use to effectively direct limited
outreach resources. These estimates were obtained either using non-probability samples (and so the
results might not be representative and one cannot measure the error in the estimates), or were obtained
via telephone surveys (which could be subject to respondents’ potential reluctance to report that their
child was not in a booster seat). (See the next section for further information on these prior estimates of
booster seat use.) What one would desire to adequately allocate limited resources for outreach programs
would be a probability-based survey in which booster seat use is obtained by observation. This is what
the NSUBS was designed to achieve. (See the design section for how the survey was designed to meet
these goals.)
Thus the National Survey of the Use of Booster Seats is being conducted to respond to the Section 14(i)
of the Transportation Recall Enhancement, Accountability, and Documentation (TREAD) Act of 2000.
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
3
3.
The Prior State of Knowledge - A History of Booster Seat
Use Estimates
3.1
What’s Difficult About Getting a Reliable Estimate of Booster Seat Use
Because of differences in reported use rates versus observed use rates (e.g., NHTSA’s Motor Vehicle
Occupant Safety Survey consistently finds reported belt use rates higher than the observed use rates in
NHTSA’s National Occupant Protection Use Survey), it is preferable to estimate booster seat use from a
survey that observes vehicles on the road, rather than one that obtains its data from telephone interviews
of drivers. However, observing booster seats presents a special challenge not encountered with, e.g., seat
belts. Namely, one type of booster seats – backless boosters – cannot be reliably observed from the
roadside. One can – and NHTSA’s National Occupant Protection Use Survey does – produce observed
estimates of high-backed booster seats of vehicles on roadways, but one cannot do the same for booster
seat use per se, i.e., the percent of children using any type of booster seat. Thus, to estimate booster use,
we are forced to take a different approach.
3.2
Estimates of Booster Seat Use Prior to NSUBS
Various approaches are possible, each with associated limitations. One can estimate use from crash data
– as Children’s Hospital of Philadelphia did using crashes of State-Farm insured vehicles. As mentioned
above, one can estimate use from telephone surveys, as NHTSA has done in its Motor Vehicle Occupant
Safety Survey (Boyle et al., 2005). However because crash-based estimates tend to underestimate use (as
drivers in crashes might disproportionately engage in risk-taking behaviors) and because of the bias of
telephone survey data discussed above, the best means of obtaining data with which to estimate booster
use would be to observe usage up close through doors and windows of stopped vehicles (with the
occupants’ consent). Doing so at a probability sample of roadways and stopping vehicles via police
checkpoints is prohibitively expensive - NHTSA estimates that such a survey would cost at least
$1,300,000 each time it is conducted. SafeKids handled the cost issues by conducting a survey at a
convenience sample of sites where vehicles containing children tend to be, such as at fast food restaurants
(Cody et al., 2002). The NSUBS does, in a sense, one better, by conducting such a survey at a probability
sample of such sites.
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
4
A Collection of Booster Seat Use Estimates
Booster Seat Use, in Percent
50%
40%
SafeKids
CHOP
MVOSS
37%
30%
27%
21%
20%
16%
10%
0%
2002
2002
2003
2003
2004
2004
2005
Notes and Sources: SafeKids estimated the percent of 4- to-7-year-olds over 40 pounds in booster seats; source: Cody et al., 2002;
CHOP estimated percent of 4- to-7-year-olds in crashes who were in boosters; source: Partners for Child Passenger Safety, 2004, 2005.
MVOSS estimated the percent of 4- to-7-year-olds in boosters at least on occasion, via telephone interviews; source: Boyle et al., 2005.
3.3
Considering an Optimal Survey
In some sense the optimal survey would be one that captured vehicles in traffic at a probability sample of
roadway sites. Setting aside the challenges of how to capture the vehicles without incurring an
unsatisfactory degree of bias (e.g., considering using police checkpoints), we feel that the number of
roadway sites one would need to collect data from would be cost-prohibitively high, given the relative
incidence of booster-age children in general roadway traffic.
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
5
4.
Sample Design
4.1
Selection of Primary Sampling Units
4.1.1 PSU Sampling Frame
The sampling frame for the first stage of the NSUBS design consists of the 50 sample Primary Sampling
Units (PSUs) used by the NOPUS in 2005, the time of the NSUBS design.
For documentation on the NOPUS PSUs and how they were selected, see Glassbrenner, September 2002.
In essence, the NOPUS PSUs, which consist of counties and groups thereof, were selected as a stratified
PPS (probability proportional to size) sample, using vehicle miles traveled (VMT) as the measure of size.
The strata used in the selection were based on four geographic regions (Northeast, Midwest, South, and
West), and whether or not the county or group of counties comprises a Metropolitan Statistical Area (or
MSA, as defined by the Office of Management and Budget) (OMB, 2005).
Note in this report the term “PSU” without further modification shall refer to the NSUBS PSUs. When
referring to the NOPUS PSUs, we shall say “NOPUS PSUs.”
4.1.2 Selection of PSUs
Sixteen PSUs were chosen from the sampling frame via the following three-step process. (The decision
to choose 16 PSUs was motivated by variance constraints - See the below section on sample size
determination for more information.)
Step 1: Two NOPUS PSUs in the frame were identified with certainty because of their population density.
An additional 22 NOPUS PSUs were selected from the remaining 48 NOPUS PSUs as an equalprobability systematic sample, with the 48 NOPUS PSUs sorted by the following three variables:
whether or not the State containing the NOPUS PSU had (in 2005 at the time of the NSUBS design) a law
requiring some children to be restrained in booster seats in at least some circumstances; whether the PSU
lies in a Metropolitan Statistical Area; and the census region. (Each of the 48 NOPUS PSUs lies entirely
within a single MSA, a single census region, and a single State or the District of Columbia).
Step 2: Fourteen NOPUS PSUs were selected from the 22 NOPUS PSUs not chosen with certainty in
Step 1 as an equal-probability systematic sample, with the 22 NOPUS PSUs sorted by the first two sort
variables from Step 1 (namely, whether or not the State containing the NOPUS PSU had a booster seat
law; and whether the NOPUS PSU lies in an MSA).
Step 3: Each of the 14 NOPUS PSUs from Step 2 and the 2 NOPUS PSUs selected with certainty in Step
1 was partitioned into county groups, where each county group consisted of a single county or two
neighboring counties. The partitioning was conducted subjectively, motivated by reducing data collection
costs in NOPUS PSUs that cover a wide geographic area. In total, 43 county groups resulted from the
partitioning of the 16 NOPUS PSUs. A single county group was selected from each of the 16 partitioned
NOPUS PSUs via PPS sampling, with the population of children under age 5 according to the 2000
Census as the measure of size. The 16 county groups resulting from these selections are the sample PSUs
for the NSUBS survey.
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
6
Thus a total of 16 PSUs was selected for the NSUBS, with each PSU consisting of a single county or two
neighboring counties that lie geographically within a NOPUS sample PSU.
Please note that consistent with our use of the phrase “PSU” in this report, the phrase “sample PSU” (e.g.,
“the 16 sample PSUs”) shall refer to the 16 NSUBS PSUs selected in Step 3 above, and not the NOPUS
sample PSUs.
The reason Step 2 was implemented instead of simply selecting 16 PSUs from the NSUBS sampling
frame via systematic sampling, is because NHTSA initially envisioned using 24 PSUs, a decision later
changed because of budget constraints. (Alternatively, and roughly equivalently, we could have
disregarded the 24-PSU result of Step 1 and re-applied Step 1 to select 2 certainty and 14 noncertainty
PSUs.)
To best ensure that the data collected at the sites reflects the actual behavior of motorists, NHTSA does
not release the locations of the 16 NSUBS (or even the NOPUS) PSUs.
Note that there is an implicit first stage of selection in the selection of the NSUBS PSUs, namely in the
selection of the NOPUS sample PSUs. As mentioned above, please see Glassbrenner, 2002, for
documentation on the selection of the NOPUS PSUs. The site selection probabilities for the NSUBS
sample sites will contain a term reflecting the NOPUS PSU selection.
The NOPUS PSUs were used to select the PSUs for the National Survey of the Use of Booster Seats,
motivated by greater comparability of the results of the two surveys. We note that the NOPUS has
adopted a new sample since the time the NSUBS PSUs were selected, and thus NHTSA may wish at
some point in the future to reselect the NSUBS sample from the current NOPUS sample for the same
reason. For more information on the current NOPUS sample, see Glassbrenner, to appear.
A note on terminology
The reader will note that the NSUBS sample design is technically a three-stage design, as the NSUBS
“PSUs” are selected in two stages, Step 3 consisting of the second stage. However as a matter of
terminology, we find it convenient to call the county groups from which the NSUBS sites were selected
“PSUs” instead of “SSUs.”
4.2
Selection of Sites Within PSUs
4.2.1 Site Sampling Frame
The sampling frame for the second stage of sampling consists of:
•
the daycare centers in the 16 sample PSUs (i.e., the 16 PSUs selected in Step 3 of Section 4.1.2),
together with
•
the recreation centers, gas stations, and restaurants in five fast food chains2 in the collection of
ZIP Codes contained in whole or in part in the 16 sample PSUs
that were found in a process described below to meet the following four restrictions:
2
The NSUBS includes among its sites restaurants in five fast food restaurant chains. In the interest of retaining
these chains in future surveys, the names of the chains (which are known to staff working on the survey) are kept
confidential in this report.
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
7
1)
2)
3)
4)
the establishment was not on a military base and not in an office building;
if the establishment was not a gas station, the establishment was not located in a shopping center;
the recreation centers did not merely contain a park, climbing wall, or senior center; and
the daycare centers were licensed for at least 20 children.
We call the above four restrictions the site sampling frame restrictions.
Formation of the Site Sampling Frame
The sampling frame was formed as follows:
•
The daycare centers in the sampling frame were compiled from State and county lists of licensed
daycare centers.
•
The gas stations were obtained by searching for all gas stations in the PSU’s ZIP Codes using the
Web sites yellowpages.com and superpages.com. The search was implemented by typing the
phrase “gas station” in the business type field and typing each of the PSU’s ZIP Codes in the ZIP
Code field of these Internet sites.
•
The fast food restaurants were obtained by searching for all such restaurants in the PSU’s ZIP
Codes using the Web sites yellowpages.com and superpages.com. The search was implemented
by typing the names of each of the five fast chains in the business type field and typing each of
the PSU’s ZIP Codes in the ZIP Code field of these Internet sites.
•
The recreation centers were obtained by compiling State and county lists of recreation centers
with the list obtained by searching for all recreation centers in the PSU’s ZIP Codes using the
Web sites yellowpages.com and superpages.com. The search was implemented by typing the
phrase “recreation center” in the business type field and typing each of the PSU’s ZIP Codes in
the ZIP Code field of these Internet sites.
Please note that the manner in which the sampling frame was formed can result in gas stations, fast food
restaurants, and recreation centers that are in the sampling frame but not in any of the sample PSUs. This
situation arises when a ZIP Code lies partly within and partly outside of a sample PSU (a phenomenon
that did occur among the 16 sample PSUs). It was impractical to address this deficiency during sampling
frame formation. We intended to remedy this deficiency after sample selection (see the section “A PostSelection Substitution We Planned to Make But Didn’t” in Section 4.2.2), but through an oversight this
was not implemented.
Pre-Selection Process to Remove Duplicates From the Frame
Note that the manner in which the sampling frame was formed can result in two or more members of the
sampling frame that identify the same establishment. This can result in a number of ways:
•
Establishments with multiple phone numbers: E.g., a recreation center with two phone numbers
might appear as two listings in an Internet search, once for each phone number (e.g., the listings
“Peoria Recreation Center, 10 Main St, Peoria IL 61602, (309) 555-1000” and “Peoria Recreation
Center, 10 Main St, Peoria IL 61602, (309) 555-1001” might appear as distinct results on
superpages.com, or one might appear on superpages.com and the other on yellowbook.com, or
one might appear on the State list of recreation centers and the other on the county list of
recreation centers).
•
Shorthand for street addresses: A fast food restaurant at 10 Main Street might appear once under
“10 Main Street” and once under “10 Main St.”
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
8
•
Establishments that have changed names: E.g., a gas station that changed names from “Bob’s
Gas” to “Steve’s Gas” might appear as two listings in an Internet search, once for each name.
•
Establishments on street corners: E.g., a McDonald’s located at the intersection of Main St. and
1st St. might appear once under Main St. and once under 1st St.
We shall call a member of the sampling frame that identifies the same establishment as another member
of the frame a duplicate (or duplicate site).
Some duplicates were identified prior to sample selection by printing the name, addresses, and phone
numbers of the establishments in the sampling frame for a given sample PSU and site type, and visually
scanning each printout to identify instances of establishments having the same address. Duplicates
identified in this manner were removed from the sampling frame.
Note that this process will not identify all instances in which a given establishment is listed multiple times
in the sampling frame. We note that we could have automated and/or refined this process of duplicate
identification using record linkage software.
We also note that we will engage in two subsequent processes for identifying duplicate sites: These
processes are described in the sections “Post-Selection Exclusions from the Sample” in Section 4.2.3 and
“Post-Selection Process to Identify Duplicates in the Frame” in Section 4.2.4.
Process to Apply the Four Site Sampling Frame Restrictions
Restrictions 1 to 3 were applied by examining the sites’ addresses. (Addresses that identified businesses
as being located in shopping centers, on military bases, or in office buildings were eliminated. E.g., a
daycare center identified as “Happy Kids Daycare, Parklawn Building” would have been eliminated.
Addresses of recreation centers that suggested the presence of merely a park, climbing wall, or senior
center were eliminated. E.g., an address identifying the center as “Rockville Climbing Wall” or “Rock
Creek Park” or “Golden Oldies Senior Center” was eliminated.) Restriction 4 was applied from licensing
information contained on the county lists.
The choice of site types and Restrictions #3 and 4 were motivated by the desire to capture large numbers
of children, particularly in the 4- to 7-year-old age range. Restriction #1 was necessitated by the ability
to access the site. Restriction #2 was motivated by the practical consideration of data collectors being
able to approach vehicles before its occupants have exited the vehicle.
There were additional sampling frame restrictions that we would have liked to apply, but were impractical
and so were effectively applied after site selection. E.g., we would have liked to restrict the sampling
frame of recreation centers to those that contain programs for children under age 12, but this was
impractical to implement as a frame restriction. Instead we implemented such restrictions through
information ascertained in phone calls to the sample sites. We will describe this process further in
Section 5.1.2.
The sampling frame was stratified by the four site types: gas stations, recreation centers, daycare centers,
and fast food restaurants. Instances of the word “strata” (or “stratum” or “stratification,” etc.) in this
report refer to this stratification.
In this report, the term site sampling frame shall refer to the sampling frame formed in this section. Thus
the site sampling frame only contains sites in the ZIP Codes of the16 selected NSUBS PSUs. As
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
9
mentioned above, in Section 4.2.3 we shall identify duplicate sites in the site sampling frame, and thus the
site sampling frame (as defined in this report) does not contain distinct members.
4.2.2 Selection of the Probability Sample of Sites
Sites were selected in a three-step process. Initially a sample of 323 sites was selected via stratified
systematic sampling (described in detail in the following). However in anticipation of businesses
declining allowing the survey to be conducted on their premises, an additional 302 sites were selected
from the remaining sampling frame. Finally, two sites were added for reasons specified below, yielding a
total of 627 sites.
Step 1: The selection of 323 sites
Initially, a target sample size of 20 sites per PSU was set, except in one PSU that was set to have 23 sites.
(See the Section 4.4 for how the target sample sizes were developed.)
The target sample size of 20 or 23 sites per PSU was allocated across strata as follows. The designated
stratum sample sizes for daycare centers and recreation centers were in all but 5 PSUs set to be 2 for each.
The numbers of daycare and recreation centers were generally significantly smaller than those of fast food
restaurants and gas stations, thus a proportional allocation would have resulted in very small sample
sizes.3 A sample size of 2 was decided upon in these cases. The remaining sample size in the PSU
(generally 16) was allocated to gas stations and fast food restaurants in proportion to their frame counts.
The stratum sample sizes in each of the 16 PSUs having been determined, the 323 sites were chosen as a
stratified systematic sample in each PSU, with the sites in a given stratum of a given PSU sorted as
follows:
•
Fast food strata in which more than 20 percent of the stratum members straddle two adjacent
counties and that have more than 25 members were sorted by chain name;
•
Gas station strata in which more than 20 percent of the stratum members straddle two adjacent
counties and that have more than 25 members were sorted in random order; and
•
All other strata were sorted by ZIP Code.
Sorting by ZIP Code ensures good geographic dispersion, and is preferred for this reason. However
because of our frame sources and sampling methods for fast food restaurants and gas stations, sorting
these strata by ZIP Codes could result in selecting an undesirably large number of sites that lie outside the
16 PSUs, and thus the alternative sorts were used.
Step 2: The selection of an additional 302 sites
The supplemental sample was formed by taking the next member in the sorted frame following each of
the selected 323 sites in the initial sample (or in the case in which the initially selected member is the last
member of a stratum, we chose the penultimate member of the stratum). The supplemental sample
contained fewer than 323 members because in some cases the “next member” was a member of the initial
sample.
3
There were a few exceptions to this. There were two noncertainty PSUs in which there were many (400 or more)
daycare centers on the frame, and a sample size of 4 or 5 was assigned in these cases. There were two other PSUs
where there was one frame unit for recreation centers in the PSU, and in this case the one frame unit was taken.
There was one PSU with no recreation centers.
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
10
Step 3: The selection of two additional sites
Two sites were inadvertently included in the sample. As we will document in Section 4.3, these sites
were treated as second-stage certainties in weighting.
We note that although Step 3 in the selection of PSUs from Section 4.1.2 involves a subjective process,
the sample of 627 sites (with the exception of the two additional sites from the previous paragraph) is a
probability sample, since the subjective process involved only the sampling frame formation, not the
selection of PSUs (or sites).
Following are the number of sites by site type in the NSUBS sample: 53 recreation centers, 75 daycare
centers, 201 fast food locations, and 298 gas stations.
We shall call the set of 627 sites resulting from the above three steps the probability sample (or for
emphasis, the NSUBS probability sample). Please note that we shall make some post-selection
refinements to the probability sample in Section 4.2.3 so that the probability sample will not consist of the
set of sites on which the survey will attempt to collect data. (See Section 4.2.3 for details.)
4.2.3 Post-Selection Refinements to the Probability Sample: Obtaining the “Refined
Sample”
In the previous section we selected 627 sites in a probabilistic manner, and called this set of sites the
probability sample. In this section, we make a number of post-selection refinements to the probability
sample. We shall call the set of sites resulting from the operations described in this section the refined
sample (or for emphasis, the NSUBS refined sample).
Post-Selection Substitutions in the Sample
One site in the probability sample was excluded via the following process. The addresses of each of the
627 sites in the probability sample were entered into the software ArcView Geographic Information
System, version 3.2 (manufactured by ESRI Corporation). This software identifies the latitude and
longitude of the addresses, through which it was discovered that one site (a gas station among the 323
sites selected in Step 1 above) was listed twice in the sample (i.e., two of the Step 1 sites had the same
latitude and longitude). Recall that the “next” member of the stratum containing this site was selected
into the probability sample in Step 2 above. The “next” member following this “Step 2” site was selected
as the substitute for the duplicate gas station.
Post-Selection Exclusions from the Sample
A total of 68 sites were excluded based on information obtained upon the data collectors visiting the site
to conduct the survey. Among these, 58 sites were found to have gone out of business or to not comply
with the four frame restriction criteria from Section 4.2.1. (E.g., a site selected as a gas station found to
not sell gas, or a site selected as one of the five fast food chains that was found to be no longer a member
of one of these chains.) In addition, 5 sites were excluded because they had no parking lots (and thus
there was no location at which to effectively conduct the survey). Finally, 5 other sites were excluded for
other reasons, such as being deemed by the data collectors as unsafe. (For breakouts of these numbers by
site type, please see the table “Business Recruitment by Site Type” in Section 5.1.2.)
As a reminder, we call the set of sites resulting from the above post-selection exclusions the refined
sample. Thus the refined sample contains 559 members (i.e., 559 sites). This is the collection of sites at
which the survey will attempt to collect data in Section 5.
Following is a depiction of the relationship between the probability sample and the refined sample:
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
11
The NSUBS Probability Sample and Refined Sample
Probability Sample
Legend
Refined Sample
one of the 68 sites excluded in Section 4.2.3 for which no substitute was chosen
(not all 68 depicted)
one of the 556 sites selected in Steps 1 or 2 of Section 4.2.2 that is not excluded in
Section 4.2.3 (not all 556 depicted)
one of the two sites selected in Step 3 of Section 4.2.2
the site excluded in Section 4.2.3 for which a substitute was chosen
the substitute site selected in Section 4.2.3
Probability
sample
Refined
sample
A Post-Selection Substitution We Planned to Make But Didn’t
Recall from Section 4.2.1 that the sampling frame contains sites not in any of the selected PSUs. This
arose only for gas stations and fast food restaurants in cases where a selected PSU contained part of a ZIP
Code. We had planned to enter location information for each of the selected fast food restaurants and gas
stations into the ArcView software. This software would have allowed us to identify the selected sites that
did not lie in any of the selected PSUs. We had planned to choose substitutes for these sites by taking the
next member in the sorted frame (or in the case in which the initially selected member is the last member
of a stratum, we would have chosen the penultimate member of the stratum). However through an
oversight this was not implemented in the 2006 sample and due to the date of discovery of the oversight.
We plan to address it in the 2008 survey.
As alluded to earlier, in the interest of data quality NHTSA does not publicly release the locations of the
observation sites or even the States in which they lie.
4.2.4 Post-Selection Process to Identify Duplicates in the Second Stage Frame
The addresses and phone numbers of the sampling frame in the selected PSUs were entered into MS
Access. MS Access identified 19 distinct duplicate sites in the frame (i.e., 19 distinct sites, each of which
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
12
had multiple occurrences on the frame), by looking for sites with the same address or same phone
number.
We note that we could have incorporated this process upon forming the sampling frame, or could have
refined this process through the use of address-matching software.
Recall from Section 4.2.1 that in this report the term site sampling frame is defined as the sampling frame
formed in Section 4.2.1. Thus the members of the sampling frame are not distinct and the process
identified in the current section identifies duplicates in the sampling frame.
4.3
Site Selection Probabilities
Because of the identification of frame duplicates in Section 4.2.4, the calculation of the site selection
probabilities is nontrivial. We shall first calculate what the site selection probabilities would have been
had the sampling frame constructed in Section 4.2.2 contained no duplication. We shall then develop an
adjustment that approximately adjusts the site selection probabilities for the frame duplication. What
shall result will be an approximate but not exact calculation of the true site selection probabilities, which
would have been prohibitively complex to calculate.
Some Notation
It is at this point that we shall need to establish some notation. We shall use the following notation
throughout the report. For the reader’s convenience, a glossary of the definitions of all mathematical
notation used in this report appears in Section 11.
Consider the NSUBS site sampling frame (which only contains sites in the 16 sample PSUs). Order the
sample PSUs and the four strata within the sample PSUs in some manner fixed for the duration of this
report.
Let 1≤ i ≤ 16 and 1≤ j ≤ 4 denote integers. Let mij denote the sample size (i.e., the number of members in
the probability sample) in the jth stratum (which we shall also call “stratum j”) of the ith PSU (i.e., PSU i).
Let m1ij (respectively, m2ij) denote the sample size for stratum j of the NSUBS PSU i in the selection of
323 sites in the probability sample selected in Step 1 of Section 4.1.2 (respectively, the 302 sites selected
in Step 2), and let m3ij take the value 1 if stratum j of PSU i contains one of the two sites selected in Step
3 of Section 4.2.1, and 0 otherwise. Thus mij = m1ij + m2ij + m3ij.
Let Mij denote the number of sites in the site sampling frame for stratum j of PSU i.
List the Mij sites in stratum j of PSU i of the site sampling frame in a manner such that:
•
the first m1ij sites in the list consist of the sites (in this stratum and PSU) that were selected in
Step 1 of Section 4.2.2, followed by
•
the m2ij sites selected in Step 2 of Section 4.2.2, followed by
•
the m3ij sites selected in Step 3 of Section 4.2.2, followed by
•
the Mij - mij sites that are not in the probability sample.
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
13
Note that the sublist described by the third bullet is nonempty only if stratum j of PSU i contains one of
the two sites selected in Step 3. Again, this sort order shall be fixed for the duration of this report.
Let 1≤ k ≤ Mij and consider site k of stratum j in PSU i of the site sampling frame.
The goal of this section is to develop the formula for the probability that site k of stratum j in PSU i is
selected in the NSUBS probability sample. Thus our formula will be defined for the members of the site
sampling frame. We shall not develop this formula for the sites in the non-sample PSUs, as the
partitioning of the NOPUS PSUs into the NSUBS PSUs involved a subjective process that was only
performed for some of the NOPUS PSUs (namely, those described in Step 3 of Section 4.1.2.)
We shall also define further notation as needed throughout the report. Notation defined anywhere in this
report appears in Section 11, together with its definition.
The Site Selection Probabilities Had There Been No Duplicates on the Frame
Had the frame formed in Section 4.2.1 contained no duplicate sites, the selection probability for site k of
stratum j in PSU i would have been:
⎧
Popi m1ij + m2ij
⎪q i δ i
TotPopi
M ij
⎪
′ := ⎨
pijk
⎪ qiδ i Popi
⎪⎩
TotPopi
for 1 ≤ k ≤ m1ij + m2ij and mij < k ≤ M ij
for m1ij + m2ij < k ≤ mij
where
qi denotes the probability of selection of the NOPUS PSU containing (the NSUBS) PSU i,
δi := 1 if the NOPUS PSU containing PSU i is one of the two certainty PSUs identified in Step 1
of Section 4.1.2, and 14/48 otherwise,
Popi := the population in 2000 of children under age 5 in PSU i,
TotPopi := the population in 2000 of children under age 5 in the NOPUS PSU containing PSU i,
Mij denotes the number of sites in the sampling frame for stratum j of PSU i,
We note that the nonunity value of the term δi reflects the combination of Steps 1 and 2 from Section
4.1.2, in which 22 PSUs were selected from 48 via systematic sampling, followed by a further systematic
subsampling of 14 PSUs from the selected 22.
Adjusting for Duplicates on the Frame: The (Approximate) Site Selection Probabilities
Recall that in Section 4.2.4 we identified 19 instances on the sampling frame in the 16 selected PSUs that
had duplicates. Thus p′ijk does not truly reflect the site selection probability, as duplicated sites had
multiple chances to be selected.
A first idea for approximating the true selection probability (in a manner that reflects the fact that
duplicates exist on the sampling frame) would be to replace the portion of p′ijk that reflects the second
stage selection with the probability that a site is selected at least once from the given PSU i and stratum j .
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
14
Letting rijk denote the number of occurrences of site k of stratum j of PSU i in the site sampling frame, we
could estimate this probability as
pk′′|ij := 1 − (1 −
′
pijk
pi
) ijk for all 1 ≤ i ≤ 16, 1 ≤ j ≤ 4, 1 ≤ k ≤ M ij
r
where pi denotes the probability that PSU i is selected (i.e., pi := qiδ i
Popi
). (We are viewing the
TotPopi
second stage selection as a Poisson selection in which members of the site sampling frame (which was
created in Section 4.2.2 and contains duplicate sites) are selected independently, each with probability
p′ijk.) (Neither of the two sites selected in Step 3 of Section 4.2.2 were duplicate sites and thus p″k|ij is
well defined.)
However the inverses of these first attempts at selection probabilities do not sum over the sample to (and
might not even be close to summing to) the frame count for stratum j of PSU i. That is,
mij
∑
k =1
ij
1
≠ M ij − ∑ (rijk −1)
′′
p k|ij
k =1
M
Thus we shall multiply the p″k|ij by a factor to arrange for this. Doing so gives that our best estimate of
the selection probability for site k of stratum j of PSU i is:
mij
pijk := pi p k′′|ij
1
∑ p ′′
l =1
l |ij
M ij
2M ij − ∑ rijk
for all 1 ≤ i ≤ 16, 1 ≤ j ≤ 4, 1 ≤ k ≤ M ij
k =1
(For the pure convenience of making pijk well defined we have utilized the same multiplicative factor pijk /
pi p″k|ij for the members of the site sampling frame that are not in the probability sample. The definition
of the site selection probabilities for these members will only matter for the purpose of estimation for the
members of the refined sample that are not in the probability sample. Alternatively, we could have
restricted the definition of pijk to those values of i,j, and k that refer to members of the refined sample.)
We shall call pijk the site selection probability for site k of stratum j of PSU i, and note that it is only
approximate. In order to truly calculate the site selection probability, we would need to calculate the
actual (and not just approximate) probability that at least one occurrence of a duplicate is selected, which
under the site selection method of stratified PPS sampling is prohibitively complex.
We define the sampling weight of a site in the site sampling frame to be the inverse of its site selection
probability. As the site selection probability is only an approximation of the true selection probability, the
sampling weight is only an approximation of the true sampling weight. Letting wijk denote the sampling
weight for site k of stratum j of PSU i, we have:
M ij
wijk =
1
=
pijk
2M ij − ∑ rijk
k =1
mij
′′ ∑
pi pijk
l =1
1
pl′′|ij
NHTSA’s National Center for Statistics and Analysis
for all 1 ≤ i ≤ 16, 1 ≤ j ≤ 4, 1 ≤ k ≤ M ij
1200 New Jersey Avenue SE., Washington, DC 20590
15
Thus note that in this section we have established an approximate site selection probability, which we call
the site selection probability (and corresponding approximate sampling weight, which we call the
sampling weight) for all sites in the site sampling frame. In particular, these terms are defined for the
members of the refined sample and the members of the probability sample.
Some Final Notes
We note that even if the site sampling frame had no duplicates, each site X that is not a daycare center and
that lies in a ZIP Code that straddles PSUs Y and Z would have two chances of being selected. Namely,
X could be selected because PSU Y was selected and X was found in the Internet search in the given ZIP
Code, or X could be selected because PSU Z was selected and X was found in the Internet search in the
given ZIP Code. We do not have the means to take this into account in the calculation of X’s selection
probability, as we do not have available the selection probabilities of the NOPUS frame PSUs that were
not selected in the NOPUS sample (and thus could not handle the case in which the ZIP Code lies partly
in a non-selected NOPUS PSU).
We note that every gas station, restaurant in the five fast food chains, daycare center, and recreation
center in the 50 States and District of Columbia that could be found by an Internet search or, for
recreation centers and daycare center, is on a county list of such establishments, and satisfies the four site
sampling frame restrictions from Section 4.2.1 had a nonzero probability of being selected for the NSUBS
probability sample. This is because the sampling frame for the NOPUS PSUs contained a partitioning of
the combined region formed by the 50 States and the District of Columbia (Glassbrenner, 2002).
The sampling frame from which the current NOPUS PSUs were chosen employs some modest sampling
frame exclusions, namely, the exclusion of 37 counties with very low vehicle miles traveled (VMT)
(Glassbrenner, to appear) Thus if the NSUBS sample is redesigned so that its PSUs are selected from the
current NOPUS PSUs (or partitionings thereof), all sites in the 50 States and DC that are outside of these
37 counties could be found by an Internet search and satisfy the four site sampling frame restrictions in
Section 4.2.1 would have a nonzero selection probability.
4.4
Sample Size Determination
In designing the sample for the NSUBS in 2005, we wished to achieve (if possible) the following cost
constraint and the variance constraints in the table “Desired Margins of Error”:
Cost Constraint
The survey shall (in 2005) cost no more than $300,000 in total survey costs, including costs for survey
preparation, training, data collection costs, quality control procedures, and production of estimates.
Desired Margins of Error
Estimate (Nationwide)
Desired Margin of Error, Using 90%
Confidence
Booster Age Children (Age 4-7)
Restraint use
Booster seat use
Child safety seat use1
Seat belt use2
NHTSA’s National Center for Statistics and Analysis
2 percentage points
2 percentage points
2 percentage points
2 percentage points
1200 New Jersey Avenue SE., Washington, DC 20590
16
Toddlers (Age 1-3)
Restraint use
Booster seat use
Child safety seat use1
Seat belt use2
1
2
2 percentage points
2 percentage points
2 percentage points
2 percentage points
Use of a rear-facing or front-facing child safety seat.
Use of a lap/shoulder or lap only seat belt.
If adequate frame information were available, one would like to approach this problem as one of
determining sample sizes (i.e., the number of PSUs, and the number of sites per PSU) to minimize
variances for a fixed cost. (This is the approach taken to design the NOPUS sample; see Glassbrenner, to
appear.)
However we did not have adequate information with which to model variances of booster seat use, and
thus the sample sizes had to be made on an intuitive basis.
Based on information from a 2005 pilot study of data collection protocols (described in Section 6.1), we
set the duration of the data collection period for the NSUBS to be 2 hours per site. Based on sample sizes
used in the National Occupant Protection Use Survey, we set the number of PSUs to be 16 and the
number of sites per PSU to be 20.
As noted in Section 4.2 we set a target sample size of 40 sites per PSU in order to ensure at least 20
participating sites per PSU.
The margins of error in the 2006 survey
The target margins of error from the table “Desired Margins of Error” do not appear to have been
achieved in the 2006 survey, which yielded the following estimated margins of error:
Achieved Margins of Error
Achieved Margin of Error, Using 90%
Confidence
Estimate (Nationwide)
Booster Age Children (Age 4-7)
Restraint use
Booster seat use
Child safety seat use1
Seat belt use2
4 percentage points
9 percentage points
NA3
6 percentage points
Toddlers (Age 1-3)
Restraint use
Booster seat use
Child safety seat use1
Seat belt use2
3 percentage points
6 percentage points
NA3
3 percentage points
1
Use of a rear-facing or front-facing child safety seat.
Use of a lap/shoulder or lap only seat belt.
3
Not computed.
2
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
17
There could be a variety of reasons why the target margins of error were not achieved. Perhaps the
intuitive setting of sample sizes (numbers of PSUs, sites per PSUs, and duration of data collection) for the
fixed cost differed greatly from the optimal determinations. In particular, the intuitive determination of
the sample sizes was based on sample sizes from NOPUS, but the NOPUS estimator differs from the
NSUBS estimator. (See Section 8 for details on these estimators.) We also note that the amount of funds
available for the survey ($300,000) was fairly small, and had we been able to collect more data, the
achieved margins of error might have been markedly lower.
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
18
5.
Data Collection Protocols
Many preliminary items are required in order to specify the instructions data collectors were given with
regards to how they were to collect data. We present these preliminary items in Sections 5.1 – 5.8,
followed by the data collection protocols in Section 5.9. Section 5.6 provides a list of all data collection
variables.
The preliminary topics covered are as follows:
• Techniques used to obtain cooperation from the data collection sites,
• The data collection schedule,
• The number, gender, and positioning of the data collectors,
• Definitions of restraint use used by the survey,
• The categories of race and ethnicity used by the survey,
• The wording of the interview questions, and
• Other assorted data collection topics and definitions.
5.1
Obtaining Site Cooperation
5.1.1 How Cooperation Was Obtained
Cooperation with recreation centers and daycare centers was obtained in advance of visiting these sites to
collect data via sending letters requesting cooperation, followed by phone calls to secure cooperation. At
times, it was also necessary to provide “hold harmless” agreements and certificates of insurance to certain
locations. In some localities, permission to use county recreation facilities was subject to the approval of
county commissioners and similar governing bodies.
Data collectors and quality control monitors approached individual fast food and gas station
establishments in person to secure cooperation. These staff received training in recruiting techniques to
try to maximize the participation rates of business establishments.
We note that in some cases, it was discovered during the process of attempting to secure cooperation that
the site was either no longer in business or had changed to an ineligible site type. An example seen of the
latter is a gas station that had changed to a car repair shop. In a few other cases it was discovered upon
visiting the site to secure cooperation (i.e., for gas stations and fast food restaurants) that collecting data at
the site could pose a safety risk to the data collectors because of vagrants congregating in the parking lot,
and these few sites were dropped from the survey. More details are provided in the next section, on site
participation rates.
5.1.2 Site Participation Rates for the 2006 Survey
In total 383 of the 559 sites in the refined sample gave permission for the survey to be conducted on their
premises.
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
19
Business Recruitment Results by Site Type
Site Type
Daycare Centers
Fast Food
Gas Stations
Recreation Centers
Total
Number of Sites in the NSUBS Probability Sample Which for the 2006 NSUBS
Survey…
Did Not Specifically
Expressly
Were
Decline
Survey
Declined
Ineligible for
Participated
Participation
But Total
in the Survey Survey
Survey
Neither
Granted
Participation Participation*
Permission
28
47
0
0
75
107
51
23
20
201
205
29
44
20
298
43
9
1
0
53
383
136
68
40
627
*These establishments were found during business recruitment to have gone out of business or have changed
business to an ineligible site type (58 cases); have no parking lot (5 cases); or were not suitable for some other
reason (5 cases).
5.2
The Data Collection Schedule
This section describes the dates and times of day that the 2006 survey was conducted, and how the data
collection schedule was determined.
Data collection for the 2006 survey was conducted during the period July 17–29, 2006. Data was
collected on all days of the week and during all daylight hours (7 a.m. to 6 p.m.).
Because children tend to be at certain site types at certain times of day, the times of day during which data
was collected varied by site type. In nearly all cases, data was collected at daycare centers in the mornings
(7 a.m. to 10 a.m.), while data at recreation centers was collected in the morning and midday (8 a.m. to 2
p.m.). At fast food restaurants, data was collected at breakfast, lunch, and dinner mealtimes (8 a.m.–10
a.m., noon–2 p.m., and 4 p.m.–6 p.m.). Gas stations were visited throughout the day (8 a.m. to 6 p.m.).
Surveys that involve visits to sites often use probabilistic algorithms to determine the schedule with which
the sites will be visited, in order to avoid bias in the times of day or days of week on which various types
of sites are visited. (These probabilistic designs are often clustered for efficient data collection, e.g., the
desired days of data collection might be subdivided into weeks, the PSUs might be randomly assigned to
weeks; the weeks subdivided into time slots for data collection, and the sites in each PSU randomly
assigned to the time slots in the assigned week.)
However the challenge of securing site cooperation in the NSUBS made it impractical to utilize a
probabilistic assignment of the data collection schedule. Rather the NSUBS sites were scheduled for
data collection as follows: Each PSU was assigned a string of consecutive days during the period July 17
– 29, during which data for the PSU would be collected. E.g., PSU 1 might have been assigned to have
its data collected during July 17 – 23. Appointments at daycare centers and recreation centers were
scheduled for times recommended by the managers of these centers as prime drop-off periods for
children. The remaining eligible time period for data collection were filled in by soliciting the cooperation
of gas station and fast food restaurant managers.
The schedule determined in the previous paragraph is called the site visitation schedule.
We note that there are conditions, described in Section 5.9, under which a site visit may be rescheduled.
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
20
5.3
Number, Gender, and Positioning of Data Collectors
5.3.1 Number and Gender of Data Collectors
Although the survey could have been conducted using a single data collector at each site, it was decided
to use two per site. At many high volume daycare and recreation centers, having two data collectors
increases the number of observations and interviews that can be obtained during the two-hour data
collection period. In addition, data collectors are required to monitor all vehicles that enter the data
collection sites (see Section 5.8.3). This would be challenging for an individual, especially when there are
numerous entrances to the parking lots. Previous experience from the NSUBS pilot study and the NOPUS
survey suggested that data collectors preferred to work in pairs rather than alone, particularly when
attempting to locate a site in an unfamiliar area and when collecting data at sites where they might be
questioned by members of the public about the authority to conduct the survey. Collecting data in teams
can also provide some measure of protection against data falsification.
The data collection was divided between the data collectors by having each data collector independently
collect data on different vehicles. Information from the NSUBS pilot study suggested that this was a
more efficient means of data collection than having the data collectors collect different survey variables
on the same vehicles. Also, having one data collector (rather than two) approach a vehicle was
considered to be potentially less threatening to the drivers.
It was also decided as a result of the NSUBS pilot study and a focus group conducted to explore views of
potential survey respondents to use female data collectors to the greatest extent possible. This was
because the information from the focus group and pilot indicated that the use of male data collectors
could decrease response rates. (Some focus group participants expressed that they would be hesitant to
talk to a male approaching their vehicle when they had their children with them.)
Two data collectors were assigned to collect all data in a given PSU. Thus there was a total of 32 data
collectors used for the 2006 survey.
5.3.2 Positioning Data Collectors at the Site to Best Collect Data
The question of where to position data collectors at the sites to best collect data turns out to be a nontrivial
one.
At some of the sites where children will be exiting the vehicle to go into the establishment (such as
daycare centers, recreation centers, and fast food restaurants where the vehicle is not going to a drive-thru
lane), children might in eager anticipation (or for other reasons) unbuckle their seat belts and car seat
harnesses after the vehicle enters the parking lot but before it parks. Since the NSUBS desires to reflect
child restraint usage on the road, we wish to observe the restraint use status before these children
unbuckle.
On the other hand it is also vastly preferable for data quality purposes to observe restraint use in a stopped
vehicle, when possible. E.g., it is difficult to record restraint use for five children in a moving vehicle,
even a slow-moving one, and it may be difficult to see whether a child in the vehicle is on a backless
booster.
Thus for sites (such as daycare centers) at which we generally expect children to exit the vehicle to go
into the establishment, it would seem best to station data collectors near the entrance to the parking lot to
collect “initial” observations for a small number of variables (specified below) that focus on capturing
child restraint use, follow the vehicle until it parks, and then conduct (with the driver’s consent) the
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
21
interview portion of the survey and make certain types of corrections (also specified below) that pertain to
restraint type but not restraint status.
For other sites, i.e., those for which we generally expect children to remain in the vehicle (such as gas
stations), it would seem best for data collectors to approach the vehicle as it is parking, and collect all
survey items (with the driver’s permission) when the vehicle is parked.
To that end, data collectors were instructed (unless this poses a safety risk or the business manager
objects) to position themselves at the following locations to begin data collection on a vehicle.
Location of Data Collectors at the Sites
Site Type
Location of Data Collectors to Collect Data on Vehicles and Occupants1
Gas Station
All data on vehicles and occupants are collected at a gas pump island.
Data on vehicles and occupants at a given fast food site are collected using one or both of
the following two paradigms:
Drive-Thru Paradigm
All data on vehicles and occupants is collected at a drive-thru lane.
Lot Entrance Paradigm
Data from a limited number of observational variables2 is collected from an entrance to
Fast Food
the parking lot, with the remainder of the variables and certain types of corrections3 to
Restaurant
observational data collected from the parked vehicle’s parking space.
Daycare
Center
Recreation
Center
The determination of which paradigms to to use is made the by data collectors upon
arriving at the site. If there is a drive-thru lane and the business manager does not object
to the survey being conducted there, then one data collector uses the Drive-Thru
Paradigm and the other uses the Lot Entrance Paradigm. Otherwise, both data collectors
use the Lot Entrance Paradigm, and if the site’s parking lot has multiple entrances, the
data collectors station themselves at different entrances.
Data from a limited number of observational variables2 is collected from an entrance to
the parking lot, with the remainder of the variables and certain types of corrections3 to
observational data collected from the parked vehicle’s parking space.
Data from a limited number of observational variables2 is collected from an entrance to
the parking lot, with the remainder of the variables and certain types of corrections3 to
observational data collected from the parked vehicle’s parking space.
1
We allowed data collectors to choose a location other than the specified location if the specified location posed a
safety risk or was not permitted by the business manager. There is a limited amount of data collected that does not
pertain to vehicles or occupants (such as the times at which data collection began and ended). See Section 5.6 for
the survey variables.
2
The variables Restraint Used and Seating Position of as many occupants who appear to be under age 13 as
possible, followed if possible (i.e., if the data collector can record accurately for this moving vehicle) by the
variables Restraint Used, Age, and Gender of the driver, followed if possible by Restraint Used, Age, Gender, and
Seating Position of other occupants. See Section 5.9 for more information.
3
Data collectors were instructed to make corrections to restraint types (e.g., whether a child is in a backless booster
seat or a seat belt) but not to restraint status (i.e., whether a child is restrained according to the definition of
“restrained” provided in the table “Definitions Used for the Survey Variable ‘Restraint Used”). See Section 5.9 for
more information.
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
22
We note that the two data collectors at a given site might be working at some distance from each other,
even when both data collectors are conducting observations at a parking lot entrance, since the site’s
parking lot might have more than one entrance.
Note that the seemingly necessary differential treatment of sites where children are expected to exit the
vehicles (versus other sites) could result in some amount of a corresponding differential population of
vehicles on which observational data is collected. As we will see in Section 5.9, data collectors collect
the observational survey variables prior to the interview survey variables. At gas station islands and in
fast food drive-thru lanes, it is possible that a potential respondent would effectively terminate the
collection of at least some, including perhaps some of the observational, survey variables (e.g., ask a data
collector what s/he is doing and to stop doing this, in which case the data collector would cease collecting
further data on the vehicle). At sites in which data collection was begun at the parking lot entrance (i.e.,
all daycare centers, all recreation centers, and the fast food restaurants using the Lot Entrance Paradigm),
potential respondents would seem to be much less likely to terminate the collection of the observational
variables, as they are driving a vehicle past the data collector at the time.
At the time of this publication we have not studied whether this possible differential population data
impacted the survey results. We only note that that there were a relatively small number of vehicles
(namely, 570 vehicles with 908 child occupants in the 2006 survey) that were observed and declined
participation in the survey when asked.
5.4
The Definition of Restraint Use Used by the Survey
The survey utilized the following definitions of restraint use.
Definitions Used for the Survey Variable “Restraint Used”
Restraint
Type
Definition Used for the National Survey of the Use of Booster Seats1
The occupant (of any age) is not in a child safety seat or booster seat. A seat belt is
across the front of his/her body and a seat belt is across his/her lap. The belt may
have slack in it and the shoulder belt may be under his/her arm.
The occupant (of any age) is not in a booster seat, has a seat belt across his/her lap,
Lap Belt
and has no seat belt across the front of his/her body.
Rear-Facing
The occupant is a child in a seat that sits on top of the vehicle seat in such a way that
Child Safety s/he faces the rear of the vehicle, and the harness straps are across his/her front. The
Seat
harness straps might be secured or not.
ForwardThe occupant is a child in a seat that sits on top of the vehicle seat in such a way that
Facing Child the occupant faces the front of the vehicle, and with harness straps that are across
Safety Seat
his/her front.
The occupant is a child in a seat with a seat back that sits on top of the vehicle seat,
High-Backed
and has a seat belt across the front of his/her body, whether lap or lap/shoulder. No
Booster Seat
harness is in use.
The occupant is a child sitting on a platform with no seat back that sits on top of the
Backless
vehicle seat, and has a seat belt across the front of his/her body, whether lap or
Booster Seat
lap/shoulder. No harness is in use.
Lap/Shoulder
Belt
1
These definitions were developed to provide characterizations that can be reliably implemented by data collectors.
They are not meant to convey any notion of what constitutes proper use or misuse of a particular restraint type.
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
23
5.5
The Categories of Race and Ethnicity Used by the Survey
In accordance with the standards for collection of race and ethnicity in Federal surveys established by the
Office of Management and Budget in the October 30, 1997, Federal Register Notice, Volume 62, Number
210, pages 58781-58790, the NSUBS uses the following categories of ethnicity:
• Hispanic or Latino,
• Neither Hispanic nor Latino,
and the following categories of race:
• White,
• Black or African-American,
• Asian,
• Native Hawaiian or Other Pacific Islander, and
• American Indian or Alaska Native.
Because the survey collects data on children and in order to not burden multiple adult occupants when
there may be restless children in the vehicle, the race/ethnicity of all occupants in a given vehicle is
obtained by asking the driver, with the data collectors being given the discretion to take into account
answers offered by other occupants. E.g., if a passenger (of any age) offers information regarding his/her
race and/or ethnicity in response to the driver being questioned about him/her, the data collector uses
his/her judgment as to whether to record the passenger’s answer or any offered by the driver.
Respondents are allowed to choose more than one race, in which case (c.f., Section 5.6) the data collector
records that the occupant has chosen more than one race, but does not record the particular races chosen.
Respondents are not allowed to choose more than one ethnicity.
5.6
The Survey Variables
The survey collects the following variables.
The survey variables fall into three categories: those pertaining to the site or data collection conditions,
those pertaining to vehicles or occupants collected by observation, and those pertaining to occupants
collected by interview.
Survey Variables Collected
Definition1
Variable
Variables Pertaining to the Data Collection Site or Data Collection Conditions (All Obtained via
Observation)
Observer
Name
The name of the data collector recording information at a given site
Booklet
Number2
The number of the booklet on which the data collector records information about the site.
Each data collector records the booklet number as “1” for the first booklet s/he uses at a
site, and increases this value by 1 for each subsequent booklet used.
Site
Identification The identification number assigned to the sample site
Number3
Whether the site is a gas station, recreation center, daycare center, or a fast food
Site Type
restaurant
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
24
Definition1
Variable
Site Name
Street
Address
Start Time
End Time
Urbanization
Weather
Conditions
Number of
Refusals
Number of
Missed
Vehicles
The name of the site, e.g., “Frank’s Gas Station”
The site’s street address, e.g., “10 Main Street, Peoria IL 61601”
The time that the data collector began collecting the data in the given booklet at the
given site, e.g., 9:12 a.m.
The time the data collector finished collecting data in the given booklet at the given site,
e.g., 11 a.m.
Whether in the consensus assessment of the data collectors the site is located in an urban,
suburban, or rural location
Whether the conditions at the time of data collection are clear, foggy, or have light rain4
The number of vehicles whose drivers declined participation in the survey
The number of passenger vehicles that appeared to have at least one occupant under age
13 for which the data collectors were not able to collect data because they were busy
collecting data for other vehicles
Variable Pertaining to Vehicles (Collected via Observation)
Vehicle
Type
Whether a vehicle appears to be a car, van/SUV, or pickup truck
Variables Pertaining to Occupants Collected via Observation
Gender
Seating
Position
Restraint
Used
Whether an occupant appears to be male or female
The seating position of the occupant5
Whether an occupant is in a rear-facing safety seat, front-facing safety seat, high-backed
booster seat, backless booster seat, or seat belt, or is unrestrained, as defined by the
Table “Definitions Used for the Survey Variable ‘Restraint Used’”
Variables Pertaining to Occupants Collected via Interview6
Whether an occupant is of Hispanic or Latino origin, or not7
Whether an occupant is White; Black or African-American; Asian; Native Hawaiian or
Other Pacific Islander; American Indian or Alaska Native; or more than one of these
Race
categories7
Weight
The weight, in pounds, of an occupant who appears to be less than 13 years old
Height
The height, in inches, of an occupant who appears to be less than 13 years old
Time Spent The number of hours (or approximate) that an occupant who appears to be under age 13
in Vehicle8
spent in the observed vehicle in the past week with that driver
Number of
The number of times in the past week that an occupant who appears to be under age 13
Visits to Gas
has visited a gas station with that driver
8
Stations
Number of The number of times in the past week that an occupant who appears to be under age 13
Visits to Fast has visited a fast food restaurant in the five restaurant chains used by the survey with that
Food8
driver9
Number of
Visits
to The number of times in the past week that an occupant who appears to be under age 13
Recreation
has visited a recreation center in the sample, with that driver
Centers8
Number of The number of times in the past week that an occupant who appears to be under age 13
Visits
to has visited a daycare center in the sample, with that driver
Ethnicity
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
25
Definition1
Variable
Daycare
Centers8
Variables Pertaining to Occupants Collected via Interview or Observation, Depending on the
Occupant’s Age
Age
For occupants who appears to be at least 13 years old: whether an occupant appears to
be 13 to 15, 16 to 24, 25 to 69, or at least 70 years old (collected via observation).
For occupants who appears at most 12 years old: the age of a child who appears to be
under age 13, in years and months (collected by asking the driver).
1
See Section 5.9 for information on the protocols used to collect these variables. The Appendix displays the data
collection forms on which the values of the survey variables were recorded.
2
The data collection forms are given to the data collectors in “booklets”. See Appendix 12.1 for a description of the
booklets and the forms they contain.
3
Each of the 559 sites in the NSUBS refined sample was given an identification number when the sample was
selected. The data collectors copy this number to the data collection form from a printed schedule given to them
listing the sites they are to visit on a given day and when they are to visit them.
4
In the interest of data quality, data collectors did not conduct the survey under other weather conditions.
5
The survey recorded up to three occupants in each of the first three rows of seats and none in any other rows (if
there were any).
6
Data collectors ask the driver whether a child occupants is under age 13 when they are not sure whether this is the
case. See the Table “Wording of Interview Questions” for the specific wording of the interview questions used to
collect these variables. At the data collector’s discretion, data collectors could also take into account answers offered
by occupants other than the driver.
7
Race/ethnicity categories as specified by the Office of Management and Budget in the October 30, 1997, Federal
Register Notice, Volume 62, Number 210, pages 58781-58790.
8
These variables were collected on the initial idea that they might be used in estimation, but ultimately they were
not used.
9
The NSUBS includes among its sites restaurants in five fast food restaurant chains. In the interest of retaining
these chains in future surveys, the names of the chains (which are known to the data collectors) are kept confidential
in this report.
We will refer to the variables listed above under “Variables Pertaining to the Data Collection Site or Data
Collection Conditions” as site variables, those (the sole variable) under “Variable Pertaining to Vehicles”
as the vehicle variables, those under “Variables Pertaining to Occupants Collected via Observation” as
observed occupant variables, and those under “Variables Pertaining to Occupants Collected via
Interviewing the Driver” as interviewed occupant variables. Note that the first three types of variables
are collected via observation, while the last is collected via interview. Note that the variable “Age” is
collected by interview or observation, depending on the age of the occupant as assessed by the data
collector collecting the interviewed occupant variables. Thus there are 12 site variables, one vehicle
variable, 4 observed occupant variables, and 10 interviewed occupant variables, for a total of 26 survey
variables (noting that the variable Age is both an observed occupant variable and an interviewed occupant
variable).
The survey recorded only one occupant per seating position. If a child occupant was sitting on the lap of
an adult, the data collectors collected data on the child and did not collect data on the adult. If more than
three persons occupied a row of a vehicle, the data collector recorded data on three of them and favored
children in determining which occupants to include.
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
26
Recording of Unknowns Strongly Discouraged
Data collectors were strongly discouraged from recording a value of “unknown”. For the nonoccupant
variables and the observed occupant variables, data collectors were instructed to choose among the values
allowed in the table above as best they could. (We note however that there were a relatively small
number of data collection forms with no values recorded for one or more of these variables, indicating
that the data collector either forgot to record the variable or was sufficiently uncertain about its value that
s/he chose to record no value.) For the interviewed occupant variables, data collectors were instructed to
record “DK” (don’t know) for any variables for which no response was provided. (We note that all item
nonresponses will be imputed prior to estimation. See Section 7.3)
5.7
The Wording of Interview Questions
The survey utilized the following wordings for its interview questions.
Because some questions pertain to children and in order not to burden multiple adult occupants when
there may be restless children in the vehicle, all questions were directed to the driver, but data collectors
were allowed to take into account responses offered by other occupants. E.g., when a data collector asks
a driver about a passenger’s race, the data collector was allowed to take into account any answer offered
by the passengers, as well as by the driver, in deciding which race to mark on the data collection form for
this occupant.
Wording of Interview Questions
Variable
Wording of Interview Question Used to Collect This Variable1
Age
Ethnicity2
Can you tell me the age of this child?
Are you (or this occupant) of Hispanic or Latino origin?
What is your race (the race of this occupant)? Please select one or more:
White
Black or African-American
Race2
Asian
Native Hawaiian or Other Pacific Islander
American Indian or Alaska Native
Weight3
Can you tell me the weight of this child?
Height3
Can you tell me the height of this child?
Time Spent in In the past week, how many minutes did this child spend in a vehicle driven by
Vehicle3
you?
Number of Visits In the past week how many times did you visit a gas station when this child was in
to Gas Stations3
a vehicle with you?
Number of Visits In the past week how many times did you go to a (Chain 1) with this child in a
to Fast Food3,4
vehicle with you? (Chain 2)? (Chain 3)? (Chain 4)? (Chain 5)?
Number of Visits
In the past week how many times did you go to a recreation center when this child
to
Recreation
was in a vehicle with you? Which one(s)?5
3
Centers
Number of Visits
In the past week, how many times did you go to a daycare center with this child in
to
Daycare
a vehicle with you? Which one(s)?5
Centers3
1
All interview questions are directed to the driver, and the data collectors were permitted to take into account
answers volunteered by other occupants, subject to the data collector’s judgment. Drivers were asked to report
answers to the questions for themselves and up to 8 other occupants of the vehicle seated in the first three rows.
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
27
2
Race/ethnicity categories as specified by the Office of Management and Budget in the October 30, 1997, Federal
Register Notice, Volume 62, Number 210, pages 58781-58790.
3
These questions were only asked concerning occupants who appeared to be under age 13.
4
The NSUBS includes among its sites restaurants in five fast food restaurant chains. In the interest of retaining
these chains in future surveys, the names of the chains (which are known to the data collectors) are kept confidential
in this report. The question as written here is worded with “Chain 1,” “Chain 2,” etc., substituted for the actual
names of the chains. The data collector uses the actual names of the chains in asking the question. The driver is
asked about the number of visits to each restaurant chain to aid him/her in his/her recollection. The data collector
records the total number of visits to each chain as the value of the variable “Number of Visits to Fast Food.”
5
The question “Which one(s)?” is used to determine whether the child visited daycare centers/recreation centers in
the survey sample. Visits to daycare centers/recreation centers not in the sample were not included in the count for
this variable.
5.8
Other Assorted Data Collection Topics and Definitions
This section presents assorted additional topics needed to describe the data collection protocols in the next
section. We cover items provided to the data collectors, the types of vehicles surveyed, the protocols used
to keep track of vehicles that do not participate in the survey, and assorted definitions.
5.8.1 Items Provided to Data Collectors
Each data collector was provided with the following items:
• Booklets of data collection forms;
• A badge with photo identification identifying him/her as an employee of WESTAT authorized to
collect data for the survey;
• A sheet (the Site Assignment Sheet) identifying the sites they are to visit, their locations, and when
they are to visit them;
• An authorization letter from NHTSA;
• A set of maps used to find their assigned data collection sites;
• A set of children’s stickers, to be used as incentives for participation;
• Copies of the NHTSA brochure A Parents Guide to Buying and Using Booster Seats in English and
Spanish; and
• A card listing the race and ethnicity categorizations used by the survey.
The Site Assignment Sheets, one for each PSU, are generated from the site visitation schedule determined
in Section 5.2. The Site Assignment Sheet for a given PSU provides the name and address of each site in
the PSU and the date and time on which it is to be visited. As there are 16 PSUs, there are 16 distinct Site
Assignment Sheets. Each data collector is provided a copy of the Site Assignment Sheet for his/her PSU.
The maps provided to the data collectors consist of both commercial street maps and customized maps
generated by WESTAT.
The badge was worn by the data collectors during data collection.
The stickers were offered to potential respondents as a small incentive to increase response rates. The
NHTSA brochure was also offered to respondents.
The authorization letter explains that WESTAT is authorized by NHTSA to collect this data and provides
contact information should a member of the public have questions about the survey. The data collectors
were instructed to offer this letter if their authority to conduct the survey is questioned (e.g., by a member
of the public or a police officer).
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
28
The data collection forms, letter of authorization, stickers, and race/ethnicity card are depicted in the
Appendix.
5.8.2 The Types of Vehicles Surveyed
The survey collects data on passenger vehicles (i.e., cars, vans, minivans, SUVs, and pickup trucks) that
have at least one occupant who appeared to be under age 13 (as many such vehicles as data collectors
were able to collect data on). This age restriction was motivated by the desire to capture data on
children, particularly 4- to 7-year-old children.
We shall call these vehicles (passenger vehicles appearing to have at least one occupant under age 13)
eligible vehicles.
5.8.3 Keeping Track of Nonparticipating Vehicles
The survey will adjust its estimates for eligible vehicles that were at the site but did not participate in the
survey.
We define two types of nonparticipating vehicles. One type we shall call refusals, which we define to be
eligible vehicles whose driver declined to answer interview questions in Step 6.6 of Section 5.9.
The other type we shall call missed vehicles, which we define to be eligible vehicles at the site during the
assigned data collection period for the site, on which no data is collected by either data collector.
One task for the data collectors (described in Section 5.9) is to record the numbers of refusals and missed
vehicles at each site. As we shall see in Section 5.9, keeping track of refusals is an easy matter. However
keeping track of missed vehicles is nontrivial.
Recall from Section 5.3.1 that the two data collectors at a given site work independently with regards to
collecting data on vehicles. Since eligible vehicles may enter and exit the site while one or both data
collectors is busy collecting data on a vehicle, and since the data collectors may be stationed at different
locations (e.g., one in the fast food drive-thru lane and one at the parking lot entrance, or at two different
parking lot entrances to the same establishment) there is no clear way to keep track of the missed vehicles
accurately.
The NSUBS allows the data collectors to decide how they can best keep track of missed vehicles at a
given site, choosing between the following two protocols:
• The data collectors choose one person between the two of them who will attempt as well as possible
to keep track (by notating a hash mark on a designated area of the booklet for each missed vehicle) of
all missed vehicles at the site, or
• Both data collectors will keep track (again by noting hash marks on the booklets) of the missed
vehicles and will communicate with each other as frequently as seems needed to ensure that each of
them is counting different missed vehicles (so that the number of missed vehicles at the site will be
the sum of their counts).
5.8.4 Miscellaneous
We note that the data collection will involve personnel called quality control monitors. There is one
quality control monitor for each pair of PSUs. This person supervises the data collectors in these two
PSUs and conducts quality control activities described in Section 6.
Finally, we note two definitions (of the terms Seating Row Limits and Suspension Conditions) that are
used in describing the data collection protocols in Section 5.9.
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
29
Seating Row Limits
Data is to be collected from no more than three occupants per row of seats in the vehicle, with no more
than two non-driving occupants in the first row, and from only the first three rows of the vehicle. If a
child is on a lap, then the child is coded for that seating position.
Suspension Conditions
• The vehicle is sufficiently out of view that observations can no longer be conducted (in the case
where the data collector is stationed at an entrance to the site’s parking lot)
• The vehicle reaches the fast food drive-thru window (if the data collector is at a fast food drive-thru)
• An occupant initiates contact of some sort with the data collector (e.g., asks “What are you doing?”,
“Do I know you?”, “Can I help you?”, engages eye contact with the data collector, etc.)
The Seating Row Limit is motivated by efficient data collection. The only passenger vehicles with a
fourth row of seats are 15-passenger vans, and a pilot study conducted before the NSUBS found only a
very small minority of 15-passenger vans at the site types. It was also found in the pilot study to be
relatively rare that more than three occupants occupy a given row of seats. Data collectors are instructed
that in the case in which more than three occupants are in a given row of seats, they should only record
the data of three of these occupants and should give preference to children under age 13 in deciding which
occupant(s) to not collect data on.
The Suspension Conditions are so named because the existence of (at least one of) the conditions will
suspend data collection until (and terminate it unless) permission is secured from the driver to pursue data
collection on his/her vehicle further.
5.9
Data Collection Protocols
We are now ready to describe the protocols that the data collectors were instructed to follow to collect the
survey data.
Each data collector conducts Steps 1 to 7 below for each site on the Site Assignment Sheet provided to
them from Section 5.8.1. If at any point during these steps the authority of the data collector to collect
data is questioned (e.g., by a member of the public or a police officer), the data collector is instructed to
offer the authorization letter from Section 5.8.1 to the inquirer.
Step 1: The data collector attempts with his/her partner to travel to the site and arrive at the time
scheduled on the Site Assignment Sheet. The data collectors use the maps provided to them in Section
5.8.1 to locate the site. If they cannot locate the site, they contact their quality control monitor, who many
in turn revise their remaining Site Visitation Schedule to decrease unutilized time. Otherwise the data
collectors proceed to Step 2.
Step 2: The data collectors jointly determine whether data can be collected at the site. Data collectors did
not collect data at a given site if any of the following occurred:
•
•
•
•
the site could not be located;
the site was found upon visit to be of an ineligible site type;
the manager on duty declined to allow the survey to be conducted at that time;
the data collectors felt uncomfortable collecting data at the site due to a matter of personal safety
(e.g., vagrants congregating in the parking lot);
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
30
•
•
weather conditions (e.g., moderate to heavy rain) precluded data collection or would have caused a
very low response rate; or
(for sites other than gas stations) the site does not have a dedicated parking lot.
In these cases, data collectors notified their data collection monitor to await further instructions. The
quality control monitor might reschedule given site or other sites as a consequence.
If data can be collected at the site, the data collectors proceed to Step 3.
Step 3: The data collectors jointly agree on values for the following survey variables, and record their
values on page 1 of the data collection form. These variables are collected via observation.
• Observer Name
• Booklet Number
• Site Identification Number
• Site Type
• Site Name
• Street Address
• Start Time
• End Time
• Urbanization
• Weather Conditions
Step 4: Data collectors identify the location at which they will position themselves at the given site in
order to collect data on vehicles and occupants, according to the instructions given in Section 5.3.2. E.g.,
if the given site is a fast food restaurant, the data collectors jointly determine whether one or both of them
will utilize the Drive-Thru Paradigm for collecting data on the vehicles they will observe.
Step 5: Data collectors decide at this time who (possibly one data collector or both) will count the missed
vehicles according to the guidance given in Section 5.8.3. The data collector(s) chosen for this task will
tally vehicles according to the procedures specified in Section 5.8.3 as well as they can while
simultaneously conducting Step 6.
Step 6: In the time remaining in the assigned two-hour time block for data collection at the site, each data
collector conducts Steps 6.1 – 6.11 repeatedly and independently of his/her partner.
Step 6.1: The data collector goes to the location identified in Step 4.
Step 6.2: The data collector identifies (according to his/her subjective assessment) the closest passenger
vehicle appearing to have at least one occupant under age 13, excluding any vehicle on which the
partner data collector is collecting data and excluding vehicles on which the data collector or partner
data collector (to the knowledge of the data collector following these instructions) has collected data.
Step 6.3: The data collector records by observation the following variables regarding the vehicle from
Step 6.2, in the order listed below, until a Suspension Condition (defined in Section 5.8.4) arises or all
of these variables have been collected:
•
the variables Restraint Used and Seating Position of all occupants appearing to be under age 13,
subject to the Seating Row Limits (defined in Section 5.8.4)
•
Restraint Used, Age, and Gender of the driver
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
31
•
Restraint Used, Seating Position, Age, and Gender of the remaining occupants, subject to the
Seating Row Limits
The data collector uses the definitions from Section 5.4 in recording the variable Restraint Used.
Step 6.4: If the data collector is situated at a parking lot entrance, s/he follows the vehicle until it parks
and the driver exits the vehicle.
Step 6.5: The data collector approaches the driver of the vehicle and recites the following text verbatim:
“Hi, my name is ____________ from Westat, a national research organization. We are conducting a
Booster Seat Survey for the National Highway Traffic Safety Administration. We would simply like to
record the restraint use of everyone in your vehicle and ask some simple questions. All your responses
and any observations I make are completely confidential.”
The data collector also offers the stickers from Section 5.8.1 to the driver for his/her participation in the
survey.
Step 6.6: If the driver declines participation, the data collector terminates the data collection on this
vehicle, records a hash mark in the “Refusals” section of the data collection form, and returns to Step 6.1.
If the driver consents to participate, the data collector proceeds to Step 6.7.
Step 6.7: The data collector collects the variable Vehicle Type.
Step 6.8: The data collector re-examines each vehicle occupant on whom data was collected in Step 6.3
from a closer standpoint to the vehicle if possible, without entering or reaching inside of the vehicle,
and make corrections to the survey data collected in Step 6.3 on these occupants on the data collection
form, with the following exception: Values of the variable Restraint Type recorded in Step 6.3 as other
than “Unrestrained” cannot be changed in this step to “Unrestrained” even that the occupant in question
appears now (or clearly is) unrestrained.
If the data collector is not sure whether one or more occupants is less than 13 years old, the data
collector asks the driver whether this is the case. If a child the data collector thought in Step 6.3 was
under 13 years turns out not to be so, the data collector revises the value of the variable Age
accordingly and records the child’s gender (as they would have done if they had guessed the age
correctly in Step 6.3). If a child whom the data collector thought in Step 6.3 was over 12 turns out not
to be so, the data collector revises the variable Age accordingly.
Step 6.9: The data collector records the variables listed but not collected in Step 6.3. E.g., if in Step
6.3, a Suspension Condition arose while the data collector was collecting data on the last occupant
appearing to be under age 13, she or he would record the remaining data for this occupant, and record
the variables listed in Step 6.3 for the driver and occupants appearing to be over age 12. As in Step 6.8,
the data collector asks the driver in any case where the data collector is not sure whether an occupant is
under age 13.
Step 6.10: The data collector obtains the values of the following variables for all occupants, subject to
the Seating Row Limits, by interviewing the driver using the questions from Section 5.7:
•
Race;
•
Ethnicity;
•
Weight (for occupants appearing to be under age 13 years, only);
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
32
•
•
•
•
•
•
•
Height (for occupants appearing to be under age 13 years, only);
Age (for occupants appearing to be under age 13 years, only);
Time Spent in Vehicle (for occupants appearing to be under age 13, only);
Number of Visits to Gas Stations (for occupants appearing to be under age 13, only);
Number of Visits to Daycare Centers (for occupants appearing to be under age 13, only);
Number of Visits to Recreation Centers (for occupants appearing to be under age 13, only); and
Number of Visits to Fast Food Restaurants (for occupants appearing to under age 13, only).
In obtaining the Race and Ethnicity variables, the data collector is instructed to show the driver the
race/ethnicity card from Section 5.8.1.
If a person other than the driver volunteers an answer to a particular question, the data collector is
instructed to use his/her judgment regarding the accuracy of the answer as to whether the non-driver’s
response appears to be more accurate. (E.g., a passenger might offer a response to the question on
his/her ethnicity and the data collector uses his/her judgment as to whether this response is more
accurate than any offered by the driver.)
Step 6.11: The data collector thanks the driver for his/her time and offers him/her the stickers and
brochure from Section 5.8.1.
Step 7: The data collector tallies the total number of refusals and records this on the data collection form.
If the data collector was recording the missed vehicles, s/he does likewise for the missed vehicles.
Note that in Step 6.8, the data collector is allowed to correct errors in the type of restraint used. E.g., if a
child appeared to be in a front-facing child seat in Step 6.3, but the data collector discovers upon the
closer inspection of Step 6.8 that the restraint is actually a high-backed booster seat, the data collector
changes the value of the variable Restraint Used for this occupant to “High-Backed Booster Seat.” What
the data collector is not permitted to do in Step 6.8 is change a value other than “Unrestrained” in Step 6.3
to “Unrestrained” in Step 6.8. The reason for this is we wish to capture as accurately as possible the
restraint use when the vehicle was on the road, and the child (or adult, in the case of seat belts) may have
unfastened the restraint between the times that Steps 6.3 and 6.8 occurred.
Note that potential survey respondents are asked, in accordance with OMB requirements, for their
voluntary participation in the survey and are assured of the confidentiality of their responses in Step 6.5.
At least some of survey variables (namely, some of the variables obtained by observation) were collected
in Step 6.3, prior to the data collectors asking for cooperation in Step 6.5. As we noted in Section 5.3.2,
this was done to capture restraint use before restraints are unfastened. However another advantage of this
approach is that it allows us to examine the response bias, and to reduce the response bias of the estimates
involving only observed data (as we will see in Chapter 8).
In order to increase response rates, the NSUBS uses a number of bilingual Spanish/English-speaking data
collectors. Drivers who could not participate in the interview portion of the survey due to other language
barriers were recorded as “Refusals” in Step 6.6.
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
33
6.
Quality Control Procedures
6.1
Pilot Testing of Data Collection Protocols
All data collection protocols were rigorously tested in a pilot study conducted at voluntarily participating
sites in Florida in 2005.
6.2
Recruitment of Field Staff
The contractor that conducted the 2006 NSUBS, WESTAT, Inc., has a field staff of thousands of data
collectors who conduct numerous surveys. For the 2006 NSUBS, WESTAT selected staff who had prior
experience conducting occupant restraint use surveys (such as the NOPUS and restraint use surveys
conducted for the Insurance Institute for Highway Safety) and who had interviewing experience.
In all, the 2006 NSUBS used 32 data collectors and 8 quality control monitors. An additional 4 backup
personnel who could serve as substitute data collectors or quality control monitors were also hired and
trained for the study. The data collectors were paired into teams of two and assigned to collect data in a
PSU relatively close to the part of the country in which they lived. Each data collection monitor was
assigned to monitor the data collection in 2 geographically proximate PSUs.
The purpose of the data collectors is to collect all survey data. The purpose of the quality control monitors
is to monitor data collection through unannounced site visits, and help as needed in securing cooperation
from sites, answering questions from data collectors during data collection, and coordinate the
rescheduling of sites for which reliable data could not be collected at their originally scheduled date and
time (e.g., due to inclement weather).
6.3
Training
Training was conducted during the period July 12–14, 2006, ending just prior to the start of data
collection on July 17, 2006.
All data collectors and quality control monitors received extensive training in protocols for interviewing
motorists and observing restraint use in a manner that is professional and as unobtrusive as possible.
Training was conducted in two components, classroom training and field training.
Classroom training comprising the following topics was transmitted via PowerPoint presentations given
by senior contractor employees who participated extensively in the pilot study and who answered
questions posed by attendees during training:
•
•
•
•
the data collection protocols from Section 5.9;
tips for and sample scripts illustrating the successful recruitment of businesses and motorists to
participate in the survey;
techniques for conducting successful interviews;
a list of questions (Frequently Asked Questions) that potential respondents might ask about the
survey, together with answers;
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
34
•
•
the authorization letters and photo identification cards; and
the laminated card for the race/ethnicity questions.
Please see the Appendix for the Frequently Asked Questions and race/ethnicity cards used for the 2006
survey.
All data collectors and quality controls monitors also participated in role playing, in which they practiced
recruiting and interviewing each other, with instructor feedback.
Actual child seats were used in training in order to facilitate training data collectors in the definition of the
restraint use survey variable.
In the field training portion of training, data collectors and quality control monitors practiced the data
collection protocols at sites near the training site at which prior cooperation had been secured, again with
feedback from the instructors.
6.4
Pre-Collection Test of Data Collectors
In order to give the highest quality results in identifying restraint use for the various types of child
restraints, training concluded with a written test on this topic given to all data collectors and quality
control monitors. In the test, test takers were to identify the restraint use of all occupants as they could
best ascertain in a series of photographs.
The results of the tests were as follows:
• The average score among all 44 test takers (the 32 data collectors, 8 quality control monitors, and 4
backup field personnel) was 95 percent.
• The average among the 8 quality control monitors was 97 percent.
• The average among the 36 data collectors and the 4 backup personnel was 92 percent.
Note that we trained 4 backup personnel, to allow for the possibility that data collectors could not conduct
the survey during the scheduled data collection period for personal reasons, did not pass the pre-collection
test, or were found to be unsuitable for the survey for whatever reason.
Of the 44 test takers, 5 scored lower than 75 percent. These people received additional training and their
data collection was monitored during the first days of data collection to ensure high quality performance.
6.5
Contact Information for Questions
Data collectors were also provided with contact information (phone numbers) for their quality control
monitor and other WESTAT staff, whom they could call with any questions that arise during data
collection.
6.6
Unannounced Site Visits
The quality control monitors conducted unannounced site visits to monitor the quality of data collection
their assigned PSUs.
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
35
7.
Data Entry, Editing, and Imputation
7.1
Data Entry
Data entry and formation of the booklet-level and page-level files
Data from the paper forms used in the survey were entered manually into a Microsoft Access database,
and contractor staff verified the correct entry of 25 percent of data forms and of all outlier values. The
records in this database are defined at the level of the data collection form. (See Section 12.1 for the data
collection forms.) That is, there is one and only one record for each form filled out by a data collector.
The booklet-level file
As there are two data collection forms, the database consists of two collections of records. One collection
is produced from the form “Booster Seat Survey Recording Form” and contains one record for each
booklet turned in by a data collector in which at least some information is recorded. Since there are at
least two booklets for each of the two data collectors and each of the 383 sites that participated in the
2006 survey, this collection contains at least 766 records and exactly 12 variables (the 12 site variables).
The page-level file
The other collection of records contains all recorded data from the untitled data collection form (i.e., the
form a copy of which appears as pages 1-20 of each booklet from Appendix 12.1.) This collection
contains one record for each page of a data collection booklet (other than the cover page) on which
information was recorded. Since a “page” records information on one vehicle from one of the two data
collectors, and data was collected on 3,489 vehicles in the 2006 survey, this collection consists of 3,489
records. As the survey records 13 occupant variables and records data on up to 9 occupants per vehicle,
this database contains 117 occupant variables. It also contains the Site Identification Number from the
cover page of the booklet containing the page, the booklet number (again from the cover page of the
booklet), the page number, and the vehicle variable, for a total of 121 variables.
These two collections of records (the booklet-level and page-level files), which existed as Microsoft
Access data tables, were imported to SAS (Statistical Analysis Software) as SAS data sets.
Data reconciliation and formation of the Master File
Since the survey estimates are occupant-related (e.g., the percentage of 4- to 7-year-old children who
were restrained in booster seats), we desire to compile the information in the above records to produce a
single file (the Master File) containing one and only one record for each occupant on which at least some
survey data was recorded and containing 28 variables (a PSU identifier, the 12 site variables in the sitelevel file above, and the page number and 14 vehicle and occupant variables from the page-level file
above).
In order to produce such a file, we have to perform some data reconciliation on the site variables when the
data collectors reported different values for a site variable (e.g. when the two data collectors at a site
reported different weather conditions), create an occupant-level file from the page-level file, and then
merge the occupant-level and booklet-level files by Site Identification Number. We perform the data
reconciliation in the basically obvious way (e.g., choosing the non-missing value when one data collector
doesn’t record a value and his/her partner does, and arbitrarily choosing one data collector’s value when
both recorded a value and they differ.
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
36
Since the 2006 survey collected data on 9,955 occupants, the Master File contains 9,955 records and (as
mentioned above) 28 variables.
7.2
Editing
No statistical editing was performed to alter the recorded values of outliers.
7.3
Imputation
The following provides a basic description of the imputation procedures used by the survey. We plan to
provide additional detail in future methodology reports.
The survey used logical imputation and hot-deck imputation, except for a few variables that were not
imputed and a few variables imputed as special cases.
7.3.1 Variables Imputed as Special Cases
The following 10 variables were imputed to have the following values.
Variables Imputed as Special Cases
Imputed Value1
Variable
Site Name
Street Address
Site Type
Start Time
End Time
The site name listed in the file created when the sample was drawn
The street address listed in the file created when the sample was drawn
The site type listed in the sampling frame
The time that data collection was scheduled to begin at the site, as listed on
the Site Assignment Sheet
The time that data collection was scheduled to end at the site, as listed on
the Site Assignment Sheet
Clear conditions
Weather Conditions
PSU
Identification
The PSU Identification Number listed in the drawn sample
Number
Site Identification Number The Site Identification Number listed in the drawn sample
Assigned in a logical manner to produce consecutive booklet numbers at
Booklet Number
each site
Assigned in a logical manner to produce consecutive page numbers in each
Page Number
booklet for each site
1
Recall from Section 7.1 that the Master File only contains a missing value for a given nonoccupant variable when
neither of the two data collectors at a given site recorded a value.
7.3.2 Variables Imputed by Logical Imputation
The following 6 variables, which are all occupant variables, were imputed via logical imputation based on
information from other occupants in the vehicle:
• Restraint Used;
• Race;
• Ethnicity;
• Age;
• Height (for occupants who appeared to be under age 13);
• Weight (for occupants who appeared to be under age 13);
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
37
E.g., if the races of some but not all occupants of a given vehicle were known than the races of the
occupants with unknown race were randomly chosen from those of the occupants with known race.
7.3.3 Variables Not Imputed
The following variables, which are the vehicle variable, 4 of the site variables, and one of the occupant
variables, were not imputed:
• Observer Name;
• Urbanization;
• Number of Refusals;
• Number of Missed Vehicles;
• Vehicle Type; and
• Seating Position.
Note that the Seating Position variable can be determined from the data collection form and so no
imputation is required.
7.3.4 Variables Imputed by Hot-Deck Imputation
All remaining variables (namely, the 6 remaining occupant variables) were imputed via hot-deck
imputation. These variables are:
• Gender;
• Time Spent in Vehicle;
• Number of Visits to Gas Stations;
• Number of Visits to Fast Food;
• Number of Visits to Recreation Centers; and
• Number of Visits to Daycare Centers.
The circumstances under which these variables were imputed
We imputed for the missing values of these survey variables except in the following case:
(NI)
For a given record R with a missing value for a given variable V among these 6 variables, we
did not impute for V in record R precisely when the values of all 10 interviewed occupant
variables were missing in record R. (This occurred precisely when the driver either declined
the entire interview portion of the survey or declined to provide any responses for a given
occupant.)
Following we describe the imputation of all variables when imputation occurred (i.e., when the condition
[NI] was not met).
Imputation for these 6 occupant variables
Missing values for these 6 occupant variables, except in the case where condition (NI) above applies,
were imputed via hot-deck imputation. Donor groups for each variable were formed using some
combination of the following variables: county group, age, sex, height, driver ethnicity, child ethnicity,
driver race, and child race.
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
38
8.
Estimation
8.1
Estimator Design
In general one can estimate at least three types of parameters that in some way measure the use of a
device in vehicles, such as booster seats. One could estimate the percent of travel time occupants spend
using the device (we shall call this the “time-based estimate”), the percent of miles occupants travel using
the device (an estimate we are not interested in for this publication and so shall not name), or one could
estimate the percent of occupants using the device at a random time. We shall call the latter the “snapshot
estimate.” One could visualize the snapshot estimator as representing what we would see if we placed an
all-seeing camera over the entire United States and took a photograph of all vehicular occupants at a
randomly chosen time.
It is easily seen that these parameters are different. For instance, consider the simple example of 2 drivers
on a block of a particular street (which we shall call Main Street) during a particular time period (e.g., 8–
10 a.m. on Monday, November 20, 2006). Driver 1 is driving on (this block of) Main Street during 8–9
a.m. and is not belted. (Driver 1 then exits Main Street at 9 a.m., not to return.) Driver 2 is on Main Street
during the entire 8–10 a.m. period and is belted the entire time. The snapshot estimate of use on Main
Street between 8–10 a.m. is 75 percent, while the time-based estimate is 66 percent (as 2 of the 3 personhours of driving were spent belted).
We naturally desire the NSUBS estimator to be consistent with that for NOPUS. The NOPUS estimator
is as follows: (For simplicity, we present only the estimator of belt use nationwide, as the subnational
estimators and estimators of other restraint types are, of course, similar.)
Belt use =
∑w F S B
∑w F S O
k
k
k
k
k
k
k
k
k
k
where k runs over the observation sites; wk denotes the inverse of the selection probability for site k; Fk
denotes the product of various nonresponse adjustment factors (see Glassbrenner, 2002, for more
information); Bk denotes the number of belted occupants observed at site k; Ok denotes the total number
of occupants observed at site k; and Sk:= Lk/sktk, where Lk (respectively, sk, tk) denotes the length of the
road segment corresponding to site k in the selection of the NOPUS sample (respectively, the estimated
speed of the vehicles observed at site k, the duration of the observation period at site k). (One might
restrict the terms Bk and Ok to occupants who appear to be over the age of 7, since this restriction,
although it is immaterial for our point on estimation, is used in NOPUS.)
Although it may not be initially obvious, the NOPUS estimator produces a snapshot of use, namely the
percent of occupants on U.S. roadways who are belted at a random (daylight, as NOPUS of course
observes during daytime) time. To see this, it may be useful to consider the analogy of balls traveling
the length of a chute. As the NOPUS observers are observing at a point on a road segment, we shall
consider balls observed at some point (for convenience, balls coming out the end of the chute). If x balls
are observed at the end of a chute of length L during t minutes, and the balls are uniformly spaced
traveling at the same, constant speed s, the number of balls on the chute at a randomly chosen time is
Lx/st. Thus the numerator of the NOPUS estimator is the number of “belted balls” on the chute (i.e.,
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
39
belted occupants on the road segment) at a random time, and the denominator similarly estimates the
(total) number of balls on the chute in a snapshot. Thus, NOPUS is a snapshot estimator. Its reference
population is the set of vehicles on all roads (subject to certain modest frame exclusions employed by the
survey) in the United States at a given point in time.
Thus we wish NSUBS to have a snapshot estimator as well. However, note that in NSUBS, there are no
“road segments” as NOPUS has, only sites. Thus there is no factor corresponding to the term “L/st” from
the previous paragraph, and the NSUBS estimator is simply as follows (expressed below using our “i-j-k”
notation for PSUs, strata, and sites from Section 4.3 and Section 11). (Again we present only the
estimator of booster seat use nationwide among 4- to 7-year-old children, as the other survey estimators
are similar.)
16
4
∑∑ ∑
Booster seat use =
i =1 j =1 k∈RefSampij
16
wijk Fijk Bijk
4
∑∑ ∑
i =1 j =1 k∈RefSampij
wijk Fijk Oijk
where RefSampij denotes the collection of members of the refined sample that are in stratum j of PSU i;
Fijk denotes the product of various adjustment factors defined below (which do not include a “travel time”
factor); Bijk denotes the number of children 4 to 7 in booster seats observed at site k in stratum j of PSU i,
and Oijk denotes the total number of children 4 to 7 observed at site k in stratum j of PSU i.
The next sections define the various adjustment factors, of which there are several, used by the NSUBS.
In Section 8.5, we will apply these to give the formula for the survey’s estimates. All adjustment factors
are defined for members of the refined sample.
Some Notation
It will be useful to establish the following notation. Let C denote a characteristic of occupants (e.g., C
might denote being of the age 4-7 years) and let R denote a restraint type (e.g., a booster seat). Let UseCR
denote the percentage of occupants restrained in restraint type R among those occupants having
characteristic C. (All survey estimates produced by the NSUBS are restraint use rates and thus have this
form.)
Some Terminology
The adjustment factors used in estimation will utilize the following terminology.
For a given occupant age 0-12 observed at a given site, we define the child to have a complete interview if
interview variables were obtained for the child in Step 6.10 of Section 5.9, a partial interview if the
variable age was collected (for the child) but at least one interview variable was not collected (for the
child), and no interview if age was not collected (for the child).
Note that a child for whom only the ethnicity (or only the ethnicity and restraint use) was collected is
considered to have no interview. The reason for this curious definition is that we will apply a simplified
nonresponse adjustment to adjust for interview data not obtained, as we found in the 2006 survey that
there were very few children on which some interview variables were collected, but not age. So, e.g.,
when estimating the restraint use of 4- to 7-year-old Hispanic children, rather than troubling to adjusting
for children on which ethnicity or age was not obtained, we shall for simplicity only adjust for the age
nonresponse.
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
40
A site is considered eligible if it is a member of the refined sample and is determined in Step 2 of Section
5.9 to be safe to collect data at, and, for sites other than gas stations, to have a dedicated parking lot. Note
that the eligibility of some members of the refined sample is unknown. A recreation center that did not
response to our requests in Section 5.1.1 to conduct the survey at their establishment is an example of a
site with unknown eligibility.
A site is considered to have participated in the survey, if it is a member of the refined sample and Step 3
of Section 5.9 was performed at the site.
Note that all sites that participated in the survey are eligible, as Step 3 of Section 5.9 is only performed at
sites found in Step 2 of Section 5.9 to be eligible.
We define a vehicle to be eligible if it is a passenger vehicle containing at least one child occupant under
the age of 13.
We define a vehicle to be observed if in Step 6.3 of Section 5.9 the data collector recorded the restraint
use of at least one occupant assessed to be under the age of 13.
8.2
Adjustment for Variation in Duration of Data Collection
The data collectors may have for a variety of reasons collected data at a given site for a period of time that
is longer or shorter than the scheduled 2 hours. If site k of stratum j in PSU i is a member of the refined
sample, we define its duration adjustment factor DurAdjijk to be
DurAdjijk :=
120
Durijk
where Durijk := the duration in minutes that the data collectors collected data from site k of stratum j of
PSU i.
8.3
Nonresponse Adjustment Factors
The NSUBS employs standard unit nonresponse adjustment, i.e. applying the ratio of total cases to known
cases. The survey has three types of response S1, S2, S3 to adjust for, each defined on a different type of
unit u1, u2, u3. Namely,
S1:= participated, u1:=eligible site
S2:= observed, u2:= vehicle
S3:= have complete or partial interview, u3:=observed child occupant
E.g., the first type of nonresponse is defined for eligible sites, and an eligible site is considered to
“respond” if it participated in the survey.
For each type of response (i.e. for each pair (St, ut) for 1 ≤ t ≤ 3), we define some number of nonresponse
cells, that together partition the refined sample. We shall denote the nonresponse cells for the pair (St, ut)
as
NRC(S)1(t ) ,..., NRC(S)Q(tt)
where Qt denotes the number of nonresponse cells for the response type pair (St, ut).
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
41
In each case (i.e., for all three values of t), the nonresponse cells consists of PSU strata (the portions
thereof in the refined sample) or unions thereof. We shall denote the members of the nonresponse cells as
the ordered triples (i,j,k) corresponding to the site k of stratum j of PSU i that lies in the nonresponse cell.
We note that the nonresponse cells for the 2006 NSUBS were defined in such a way that Q1 = 52, Q2 =
559 (i.e., the size of the refined sample), and Q3 = 54. (That is, the nonresponse cells used to adjust for
unobserved vehicles are precisely the members of the refined sample.)
Due to the hierarchical nature of the unit types u1, u2, u3, we shall need to define the nonresponse
adjustment factors NRAdj(S) (ts ) for 1 ≤ s ≤ Qt recursively.
(t )
For each 1 ≤ t ≤ 3, 1 ≤ i ≤ 16, 1 ≤ j ≤ 4, and each value of k for which site k∈RefSampij, let Tot ijk
(t )
denote the number of
denote the number of units of type ut at site k of stratum j of PSU i, and let Resp ijk
units of type ut at site k of stratum j of PSU i for which we have a response for St.
E.g., if j=2 denotes the gas station stratum and the 3rd gas station in the sampling frame for PSU 1 is in the
(1)
refined sample, then Tot 123
equals 1 or 0 depending on whether this gas station was eligible (i.e., was
(1)
equals 1 or 0 depending on
considered by the data collectors to be safe to collect data at), while Resp123
whether this gas station participated in the survey.
(t )
(t )
and Resp ijk
are zero.
Note that if t>1 and if site k did not participate in the survey, then both Tot ijk
The case of t=1 will be a special case, because we can only estimate the numerator of the nonresponse
adjustment factors, as the eligibility of some sites is unknown. We define the nonresponse factors for t=1
(i.e., for S1 = participate and u1=eligible site) to be:
∑
NRAdj (S ) (1)
s :=
w
ijk
(i, j ,k )∈NRC ( S )1s ∩ Elig
+ es
∑
∑
w
ijk
(i, j ,k )∈NRC ( S )1s ∩UnknownEli g
(1)
Resp ijk
wijk
for 1 ≤ s ≤ Q1
(i, j ,k )∈NRC ( S )1s ∩ Elig
where Elig:={(i,j,k): site k in stratum j of PSU i is known to be eligible}, UnknownElig:={(i,j,k): the
eligibility of site k in stratum j of PSU i is unknown}, and
es :=
∑
w
∑
wijk
ijk
(i, j,k )∈NRC (S )1s ∩Elig
for 1 ≤ s ≤ Q1
(i, j,k )∈NRC (S )1s \UnknownElig
Since vehicles are a smaller unit than sites, we shall need to incorporate the nonresponse factors for t=1
into those for t=2. That is, we define the nonresponse factors for t=2 (i.e., for S2 = observed and
u2=vehicle) to be:
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
42
∑
NRAdj (S ) (s2 ) :=
(2)
Tot ijk
NRAdj (S ) (s1)1 (i , j ,k ) w ijk
(i , j ,k )∈NRC ( S ) 2s
∑
(2)
Resp ijk
NRAdj (S ) (1)
s1 (i , j ,k ) w ijk
for 1 ≤ s ≤ Q 2
(i , j ,k )∈NRC ( S ) 2s
where for 1 ≤ v ≤ 3, sv(i,j,k) denotes the value a for which (i,j,k) ∈ NRC ( S )(av )
Similarly the nonresponse factors for t=3 incorporates those for t=1 and t=2. That is, we define the
nonresponse factors for t=3 (i.e., for S3 = have complete or partial interview and u3=observed child
occupant) to be:
2
∑
NRAdj ( S )
(3)
s
:=
(i ,
(3)
Tot ijk
(v)
sv ( i , j , k )
wijk
for 1 ≤ s ≤ Q3
2
∑
Resp
(3)
ijk
∏ NRAdj (S )
v =1
( i , j , k )∈NRC ( S ) 3s
8.4
∏ NRAdj(S )
v =1
j , k )∈NRC ( S ) 3s
(v)
sv ( i , j , k )
wijk
Weight Trimming
Consider the product of the sampling weight wijk together with the adjustment factors from Sections 8.2
and 8.3, i.e. consider:
′ := wijk
wijk
3
∏ NRAdj (S )
v =1
(v)
sv ( i , j , k )
DurAdjijk
Considering this as a “weight”, we shall define two associated trimmed weights, one used when the
characteristic C involves at least one interview variable and one when it doesn’t.
To define the two trimmed trimming factors, define
w OBS :=
1
| RefSamp |
w INT :=
where RefSamp :=
16
∑
ObsChild ijk
( i , j , k )∈RefSamp
1
| RefSamp |
∑
′
wijk
NRAdj(S) (3)
s3 ( i , j , k )
, and
′
IntChild ijk wijk
(i, j,k )∈RefSamp
4
UU RefSamp
ij
denotes the refined sample from Section 4.2.3, and for each i,j,k
i =1 j =1
for which k∈RefSampij, ObsChildijk (respectively, IntChildijk) denotes the number of children observed
(respectively, have a complete or partial interview) at site k of stratum j of PSU i. (Note that if site k did
not participate in the survey, then both ObsChildijk and IntChildijk are zero.)
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
43
We shall trim the “weight” back to 4.5 times the corresponding weighted mean of these “weights”.
Namely, we define the weight-trimming factors TrimOBSijk and TrimOBSijk as:
8.5
OBS
Trimijk
⎧ 1, if wijk
′ ≤ 4.5 w OBS
⎪
:= ⎨ 4.5 w OBS
, otherwise
⎪ w′
ijk
⎩
TrimijkINT
⎧ 1, if wijk
′ ≤ 4.5 w INT
⎪
:= ⎨ 4.5 w INT
, otherwise
⎪ w′
ijk
⎩
The Estimation Formula
We are now ready to write down the estimation formula. Let C denote a characteristic of occupants (e.g.,
C might denote being of age 4 to 7) and let R denote a restraint type (e.g., a booster seat). We shall next
provide the formula for the survey’s estimate UseCR of the percentage of occupants restrained in restraint
type R among those occupants having characteristic C. (All survey estimates produced by the NSUBS
are restraint use rates and thus have this form.)
The NSUBS estimator of UseCR is as follows:
Estimates Involving Interview Data
If C involves at least one variable obtained by interview, then we estimate UseCR by the following
formula:
UseCR :=
3
∑
TrimijkINT wijk ∏ NRAdj(S)(v)
s v (i, j,k ) DurAdjijk Bijk
∑
TrimijkINT wijk ∏ NRAdj ( S ) (svv )(i, j,k ) DurAdjijk Oijk
(i, j,k )∈Re fSamp
(i, j,k )∈Re fSamp
v =1
3
v =1
where for each (i,j,k) in RefSamp, Bijk denotes the number of occupants of characteristic C restrained in
restraint type R at site k of stratum j in PSU i, and Oijk denotes the number of occupants of characteristic
C at site k of stratum j in PSU i.
Examples of survey estimates that would be computed using this formula would be:
•
booster seat use among 4- to 7-year-old children (since the ages of children are obtained by
interview),
•
the use rate for front-facing child safety seats among children weighing 20 to 40 pounds who
appear to be under age 13 (as weight is obtained by interview), and
•
restraint use among Hispanic children (as ethnicity is obtained by interview).
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
44
Estimates Involving No Interview Data
For all other values of C (i.e., when all variables involved in C are obtained by observation), then we
estimate UseCR to be:
UseCR :=
2
∑
OBS
Trimijk
wijk ∏ NRAdj(S)(svv )(i, j,k ) DurAdjijk Bijk
∑
OBS
Trimijk
wijk ∏ NRAdj(S) (v)
s v (i, j,k ) DurAdjijk Oijk
(i, j,k )∈Re fSamp
(i, j,k )∈Re fSamp
v =1
2
v =1
Examples of survey estimates that would be computed using this formula would be:
•
restraint use among children who appear to be under the age of 13,
•
belt use among 16- to 24-year-olds, and
•
restraint use among children taken to gas stations who appear to be under age 13 (since restraint
use, the ages of occupants who appear to be over age 12, and whether or not an occupant is under
13 are assessed by observation).
8.6
Estimates Computed
The survey computes estimates of restraint use by a variety of characteristics derived from the survey
variables listed in Section 5.4. For instance the survey estimates restraint use by age and restraint type,
and by height and restraint type. The following publications present the major estimates from the 2006
survey: Glassbrenner and Ye, DOT HS 810 796, August 2007; Glassbrenner and Ye, DOT HS 810 797,
August 2007; Glassbrenner and Ye, DOT HS 810 798, August 2007.
8.7
Definitions of Categories Used in Estimates
Although the NSUBS collects children’s individual ages, heights, and weights, we combine these results
into categories in order to produce reliable estimates.
Age categories
The NSUBS uses the following age categories: 0, 1-3, 4-7, 8-12, 13-15, 16-24, 25-69, and 70 and above.
The choice of these age groups is motivated by consistency with the NOPUS survey, which uses the age
groups 0, 1-3, 4-7, 8-12, 13-15, 16-24, 25-69, and 70 and above, combined with taking into account that
the NSUBS collects interview data on children ages 0-12.
Height and weight categories
The NSUBS uses the following height categories: under 36 inches tall, 37-53 inches, 54-56 inches, and 57
inches or taller. The survey uses the weight categories 0-19 pounds, 20-40 pounds, 41-60 pounds, and 61
pounds or heavier. These categories were chosen because they are used in NHTSA’s recommendation
for the choice of restraint use for children.
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
45
Regional categories
The 16 PSUs selected in the NSUBS constitute a probability sample of PSUs (counties and groups
thereof) in the United States. The data is not sufficient to produce State-by-State results. However
NSUBS can and does produce regional estimates using the following categories:
Northeast:
Midwest:
South:
West:
ME, VT, NH, MA, RI, CT, NY, PA, NJ
MI, OH, IN, IL, WI, MN, IA, MO, KS, NE, SD, ND
WV, MD, DE, VA, KY, TN, NC, SC, GA, FL, AL, MS, AR, LA, OK, TX, DC
AK, WA, OR, CA, NV, ID, UT, AZ, NM, CO, WY, MT, HI
These definitions of the four NSUBS regions are the same regional definitions used in the NOPUS. The
NSUBS regional categories were chosen to be the same as the NOPUS categories for the purpose of
consistency.
Time of day and day of week categories
The NSUBS uses the following day of week and time of day categories, which are the same used for the
2006 NOPUS:
Weekday Rush Hour: 8-10 a.m. and 3:30-6 p.m. on Monday-Friday
Weekday Outside of Rush Hour: 10 a.m.–3:30 p.m. on Monday-Friday
Weekend: 8 am–6 pm on Saturday and Sunday
8.8
A Note on the Race/Ethnicity Estimates
When computing estimates by race and/or race/ethnicity, multiracial occupants are excluded (i.e., we did
not impute a single race for persons reporting they are multiracial). Also we had to collapse some race
and race/ethnicity categories in order to comply with NHTSA standards for reliability in publishing
estimates. (See Section 10 for more information on these standards.) A common situation in which
collapsing categories was necessary was in estimating use rates among Hispanic non-Whites.
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
46
9.
Variance Estimation
This section documents the variance estimation procedures utilized by the survey, without providing the
motivation for these procedures. We plan to provide the motivation for these procedures in a subsequent
methodology report.
Two methods for calculating variances were employed, depending on whether the estimator whose
variance is being estimated is based on relatively few observations (and thus is such that direct estimates
of variances are not reliable).
Variance estimates for estimates in the following table were computed directly via jackknife variance
estimation:
Estimates Whose Variances Were Estimated Directly
Restraint Use Among Children Age 0-12
Restraint Use Among Children Age 0-12 in Passenger Cars
Restraint Use Among Children Age 0-12 in Vans and Sport Utility Vehicles
Restraint Use Among Children Age 0-12 in Pickup Trucks
Restraint Use Among Children Age 0-12 in the Front Seat
Restraint Use Among Children Age 0-12 in the Second Row of Seats
Restraint Use Among Children Age 0-12 in Pickup Trucks in the Third Row of Seats
Restraint Use Among Children Age 0-12 Driven by a Driver Age 16 to 24
Restraint Use Among Children Age 0-12 Driven by a Driver Age 25 to 69
Restraint Use Among Children Age 0-12 Driven by a Driver Age 70 or Older3
Restraint Use Among Children Age 0-12 Driven by a Driver Whose Ethnicity Is Hispanic or Latino
and Whose Race Is American Indian or Alaska Native3
Restraint Use Among Children Age 0-12 Driven by a Driver Whose Ethnicity Is Hispanic or Latino
and Whose Race Is Asian3
Restraint Use Among Children Age 0-12 Driven by a Driver Whose Ethnicity Is Hispanic or Latino
and Whose Race Is Black or African-American3
Restraint Use Among Children Age 0-12 Driven by a Driver Whose Ethnicity Is Hispanic or Latino
and Whose Race Is Native Hawaiian or Other Pacific Islander3
Restraint Use Among Children Age 0-12 Driven by a Driver Whose Ethnicity Is Hispanic or Latino
and Whose Race Is White
Restraint Use Among Children Age 0-12 Driven by a Driver Whose Ethnicity Is not Hispanic nor
Latino and Whose Race Is American Indian or Alaska Native3
Restraint Use Among Children Age 0-12 Driven by a Driver Whose Ethnicity Is not Hispanic nor
Latino and Whose Race Is Asian3
Restraint Use Among Children Age 0-12 Driven by a Driver Whose Ethnicity Is not Hispanic nor
Latino and Whose Race Is Black or African-American
Restraint Use Among Children Age 0-12 Driven by a Driver Whose Ethnicity Is not Hispanic nor
Latino and Whose Race Is Native Hawaiian or Other Pacific Islander3
Restraint Use Among Children Age 0-12 Driven by a Driver Whose Ethnicity Is not Hispanic nor
Latino and Whose Race Is White
Restraint Use Among Children Age 0-12 Driven by a Male Driver
Restraint Use Among Children Age 0-12 Driven by a Female Driver
Restraint Use Among Children Age 0-12 Driven by a Belted Driver
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
47
Restraint Use Among Children Age 0-12 Driven by an Unbelted Driver
Restraint Use Among Children Age 0-12 Months3
Restraint Use Among Children Age 1-3 Years
Restraint Use Among Children Age 4-7 Years
Restraint Use Among Children Age 8-12 Years
Restraint Use Among Boys Age 0-12
Restraint Use Among Girls Age 0-12
Restraint Use Among Children Age 0-12 Whose Ethnicity Is Hispanic or Latino and Whose Race Is
American Indian or Alaska Native3
Restraint Use Among Children Age 0-12 Whose Ethnicity Is Hispanic or Latino and Whose Race Is
Asian3
Restraint Use Among Children Age 0-12 Whose Ethnicity Is Hispanic or Latino and Whose Race Is
Black or African-American3
Restraint Use Among Children Age 0-12 Whose Ethnicity Is Hispanic or Latino and Whose Race Is
Native Hawaiian or Other Pacific Islander3
Restraint Use Among Children Age 0-12 Whose Ethnicity Is Hispanic or Latino and Whose Race Is
White
Restraint Use Among Children Age 0-12 Whose Ethnicity Is not Hispanic nor Latino and Whose Race
Is American Indian or Alaska Native3
Restraint Use Among Children Age 0-12 Whose Ethnicity Is not Hispanic nor Latino and Whose Race
Is Asian3
Restraint Use Among Children Age 0-12 Whose Ethnicity Is not Hispanic nor Latino and Whose Race
Is Black or African-American
Restraint Use Among Children Age 0-12 Whose Ethnicity Is not Hispanic nor Latino and Whose Race
is Native Hawaiian or Other Pacific Islander3
Restraint Use Among Children Age 0-12 Whose Ethnicity Is not Hispanic nor Latino and Whose Race
Is White
Restraint Use Among Children Age 0-12 Whose Height Is Under 36 Inches
Restraint Use Among Children Age 0-12 Whose Height Is 37-53 Inches
Restraint Use Among Children Age 0-12 Whose Height Is 54-56 Inches
Restraint Use Among Children Age 0-12 Whose Height Is 57 Inches or More
Restraint Use Among Children Age 0-12 Whose Weight Is Under 19 Pounds3
Restraint Use Among Children Age 0-12 Whose Weight Is 20-40 Pounds
Restraint Use Among Children Age 0-12 Whose Weight Is 41-60 Pounds
Restraint Use Among Children Age 0-12 Whose Weight Is At Least 61 Pounds
Restraint Use Among Children Age 0-12 in Vehicles in Light Precipitation
Restraint Use Among Children Age 0-12 in Vehicles in Fog3
Restraint Use Among Children Age 0-12 in Vehicles in Clear Weather Conditions
Restraint Use Among Children Age 0-12 in Vehicles in the Northeast1
Restraint Use Among Children Age 0-12 in Vehicles in the Midwest1
Restraint Use Among Children Age 0-12 in Vehicles in the South1
Restraint Use Among Children Age 0-12 in Vehicles in the West1
Restraint Use Among Children Age 0-12 in Vehicles in Urban Areas
Restraint Use Among Children Age 0-12 in Vehicles in Suburban Areas
Restraint Use Among Children Age 0-12 in Vehicles in Rural Areas
Restraint Use Among Children Age 0-12 in Vehicles During Weekday Rush Hour2
Restraint Use Among Children Age 0-12 in Vehicles on Weekdays Outside of Rush Hour2
Restraint Use Among Children Age 0-12 in Vehicles on Weekdays
Restraint Use Among Children Age 0-12 in Vehicles on Weekends
Restraint Use Among Children Age 0-12 in Vehicles at Gas Stations
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
48
Restraint Use Among Children Age 0-12 in Vehicles at Fast Food Restaurants
Restraint Use Among Children Age 0-12 in Vehicles at Daycare Centers
Restraint Use Among Children Age 0-12 in Vehicles at Recreation Centers
1
See Section 8.3 for the definitions of the NSUBS regional categories of Northwest, Midwest, South, and West.
See Section 8.3 for the definition of rush hour.
3
These variables had fewer that 200 observations in the 2006 survey.
2
For these estimates we also computed the within- and between-PSU variances (again directly, through
jackknife variance estimation), and we calculated the average of the ratios of the total variance to the
within-PSU variance for all estimates in this table having 200 or more observations.
R: =
1 48 Var(Yi )
∑
48 i =1 WVar(Yi )
(1)
where Yi denotes the ith member of the table “Estimates Whose Variances Were Computed Directly” that
have 200 or more observations, and for a random variable Y defined on the NSUBS sample, Var(Y)
(respectively, WVar(Y)) denotes the jackknife-calculated estimate of the variance of Y (respectively, the
jackknife-calculated estimate of the within-PSU variance of Y ).
The value of R from the 2006 survey data was 3.12.
The variances of all other estimates were computed by calculating the within-PSU variance of the
variable via jackknife and multiplying by the ratio from (1), i.e.:
Var(Y): = R × WVar(Y)
for each estimate Y other than those in the table “Estimates Whose Variances Were Estimated Directly”.
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
49
10. Rules for Suppressing Estimates in Publications
In order not to publish estimates that are not sufficiently reliable, NHTSA employs the following
suppression rule for the NSUBS:
NSUBS Suppression Rule
Use estimates whose numerator is based on fewer than 5 persons observed, whose denominator is based
on fewer than 30 persons observed, or that are not statistically different from 0 percent use (i.e. the
standard error is at least half the point estimate) are to be suppressed. These should be reported as “NA”
in publications, and any related estimates (i.e., change in use and confidence estimates) should also be
suppressed.
This the same rule used for the NOPUS survey.
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
50
11. Glossary of Terms
The following records terms and acronyms that are defined in this report:
Complete interview: For a given occupant up to age 12 observed at a given site, we define the child to
have a complete interview if values of all interview variables were obtained for the child in Step 6.10 of
Section 5.9.
Duplicate (or duplicate site): a member of the site sampling frame formed in Section 4.2.1 identifying the
same establishment as another member of this frame.
Eligible site: a site in the refined sample determined in Step 2 of Section 5.9 to be safe to collect data at,
and, for sites other than gas stations, to have a dedicated parking lot.
Eligible site type: the following 4 types of establishments: fast food restaurant, gas station, daycare
center, recreation center.
Eligible vehicle: a passenger vehicle containing at least one child occupant under age 13.
Interviewed occupant variables (or for short, interview variables): The 10 survey variables that pertain to
occupants and whose values are obtained by interview for at least some occupants, namely Ethnicity,
Race, Weight, Height, Time Spent in Vehicle, Number of Visits to Gas Stations, Number of Visits to Fast
Food, Number of Visits to Recreation Centers, Number of Visits to Daycare Centers, and Age.
Interview variables (or interviewed occupant variables): The 10 survey variables that pertain to occupants
and whose values are obtained by interview for at least some occupants, namely Ethnicity, Race, Weight,
Height, Time Spent in Vehicle, Number of Visits to Gas Stations, Number of Visits to Fast Food, Number
of Visits to Recreation Centers, Number of Visits to Daycare Centers, and Age.
Master File: The file created in Section 7.1, which contains one record for each occupant observed in the
survey and all 28 survey variables.
Missed vehicle: An eligible vehicle observed according to the protocols in Section 5.8.3 on which no data
is collected by either data collector in Step 6 of Section 5.9.
No interview: For a given occupant up to age 12 observed at a given site, we define the child to have no
interview if the no value for the variable Age was obtained for the child in Step 6.10 of Section 5.9, or the
driver declined to give an interview in Step 6.6 of Section 5.9.
NOPUS: The National Occupant Protection Use Survey.
NOPUS PSU: a primary sampling unit for the NOPUS.
NSUBS: The National Survey of the Use of Booster Seats.
The NSUBS probability sample (or for short, the probability sample): The collection of 627 gas stations,
recreation centers, daycare centers, and fast food restaurants selected in Steps 1-3 of Section 4.2.2
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
51
The NSUBS refined sample (or for short, the refined sample): The collection of 559 gas stations,
recreation centers, daycare centers, and fast food restaurants resulting from Section 4.2.3.
Observed occupant variables: The 4 survey variables that pertain to occupants and whose values are
obtained by observation for at least some occupants, namely Gender, Seating Position, Restraint Used,
and Age.
Observed vehicle: a vehicle on which a data collector recorded the restraint use of at least one occupant
assessed to be under the age of 13 in Step 6.3 of Section 5.9.
Partial interview: For a given occupant up to age 12 observed at a given site, we define the child to have a
partial interview if a value for the variable Age was obtained for the child in Step 6.10 of Section 5.9, and
there was at least one interview variable whose value was not obtained for the child in Step 6.10 of
Section 5.9.
Participating site: A member of the refined sample for which Step 3 of Section 5.9 was performed.
Passenger vehicle: a passenger car, van, sport utility vehicle, or pickup truck.
The probability sample (or for emphasis, the NSUBS probability sample): The collection of 627 gas
stations, recreation centers, daycare centers, and fast food restaurants selected in Steps 1-3 of Section
4.2.2
PSU, or NSUBS PSU: a primary sampling unit for the NSUBS.
The refined sample (or for emphasis, the NSUBS refined sample): The collection of 559 gas stations,
recreation centers, daycare centers, and fast food restaurants resulting from Section 4.2.3.
Refusal: An eligible vehicle whose driver declined to answer interview questions in Step 6.6 of Section
5.9.
Sample PSU (e.g., “the 16 sample PSUs”): one of the 16 NSUBS PSUs selected in Step 3 of Section
4.1.2.
Sampling weight: the inverse of the site selection probability.
Seating Row Limits: The condition, used in Section 5.9, that data is to be collected from no more than
three occupants per row of seats in the vehicle, with no more than two non-driving occupants in the first
row, and from only the first three rows of the vehicle.
Site Assignment Sheet: A sheet identifying the sites the data collectors are to visit, their locations, and
when they are to visit them
Site sampling frame: the sampling frame of gas stations, recreation centers, daycare centers, and fast food
restaurants in the 16 sample PSUs formed in Section 4.2.1.
Site sampling frame restrictions: The following set of four restrictions, which apply to establishments that
are gas stations, fast food restaurants, daycare centers, and recreation centers:
1) the establishment was not on a military base and not in an office building;
2) if the establishment was not a gas station, the establishment was not located in a shopping center;
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
52
3) the recreation centers did not merely contain a park, climbing wall, or senior center; and
4) the daycare centers were licensed for at least 20 children.
Site selection probability: the approximation to the site selection probability calculated in Section 4.3.
Site variables: The 12 survey variables that pertain to the data collection site or data collection conditions,
namely Observer Name, Booklet Number, Site Identification Number, Site Type, Site Name, Street
Address, Start Time, End Time, Urbanization, Weather Conditions, Number of Refusals, Number of
Missed Vehicles.
Site visitation schedule: The schedule determined in Section 5.2, which specifies the dates and times at
which the survey is to be conducted at each member of the refined sample.
The strata: The stratification of the NSUBS sampling frame by the four site types: gas stations, recreation
centers, daycare centers, and fast food restaurants.
Suspension Conditions: The following three conditions, which are used in Section 5.9:
• the vehicle is sufficiently out of view that observations can no longer be conducted (in the case
where the data collector is stationed at an entrance to the site’s parking lot);
• the vehicle reaches the fast food drive-thru window (if the data collector is at a fast food drivethru)
• an occupant initiates contact of some sort with the data collector (e.g. asks “What are you
doing?”, “Do I know you?”, “Can I help you?”, engages eye contact with the data collector, etc).
SUV: sport utility vehicle.
Vehicle variable: The sole survey variable that pertains to vehicles, namely Vehicle Type.
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
53
12. Glossary of Notation
For convenience, we record in the following notation that is defined in this report.
i: denotes a sample PSU, and takes an integer value between 1 and 16
j: denotes a stratum, and takes an integer value between 1 and 4
mij: denote the number of members in the probability sample in the jth stratum of the ith PSU.
m1ij (respectively, m2ij): denotes the number of members of the probability sample for stratum j of the
NSUBS PSU i in the selection of 323 sites in the probability sample selected in Step 1 of Section 4.1.2
(respectively, the 302 sites selected in Step 2)
m3ij: takes the value 1 if stratum j of PSU i contains one of the two sites selected in Step 3 of Section
4.2.1, and 0 otherwise.
Mij: denotes the number of sites in the site sampling frame for stratum j of PSU i.
The members of the site sampling frame are sorted as follows. List the Mij sites in stratum j of PSU i of
the site sampling frame in a manner such that:
•
the first m1ij sites in the list consist of the sites (in this stratum and PSU) that were selected in
Step 1 of Section 4.2.2, followed by
•
the m2ij sites selected in Step 2 of Section 4.2.2, followed by
•
the m3ij sites selected in Step 3 of Section 4.2.2, followed by
•
the Mij - mij sites that are not in the probability sample.
qi: denotes the probability of selection of the NOPUS PSU containing (the NSUBS) PSU i,
δi: takes the value 1 if the NOPUS PSU containing PSU i is one of the two certainty PSUs identified in
Step 1 of Section 4.1.2, and 14/48 otherwise,
Popi: denotes the population in 2000 of children under age 5 in PSU i,
TotPopi: denotes the population in 2000 of children under age 5 in the NOPUS PSU containing PSU i,
⎧
Popi m1ij + m2ij
⎪q i δ i
TotPopi
M ij
⎪
′ := ⎨
pijk
⎪ qiδ i Popi
⎪⎩
TotPopi
for 1 ≤ k ≤ m1ij + m2ij and mij < k ≤ M ij
for m1ij + m2ij < k ≤ mij
rijk: denotes the number of occurrences of site k of stratum j of PSU i in the site sampling frame.
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
54
pi := qiδ i
Popi
TotPopi
pk′′|ij := 1− (1−
′
pijk
pi
) ijk for all 1 ≤ i ≤ 16, 1 ≤ j ≤ 4, 1 ≤ k ≤ M ij
r
mij
pijk := p i p k′′|ij
1
∑ p ′′
l =1
l |ij
M ij
2M ij − ∑ rijk
for all 1 ≤ i ≤ 16, 1 ≤ j ≤ 4, 1 ≤ k ≤ M ij
k =1
M ij
wijk
1
=
=
pijk
2M ij − ∑ rijk
k =1
mij
′′ ∑
pi pijk
l =1
1
pl′′|ij
for all 1 ≤ i ≤ 16, 1 ≤ j ≤ 4, 1 ≤ k ≤ M ij
RefSampij: denotes the collection of members of the refined sample that are in stratum j of PSU i
Durijk := the duration in minutes that the data collectors collected data from site k of stratum j of PSU i.
DurAdjijk :=
120
, defined for all i, j, and for k∈RefSampij:
Durijk
S1, S2, S3: denote three types of response, each defined on a different type of unit u1, u2, u3.
Qt: denotes the number of nonresponse cells for the response type pair (St, ut).
NRC(S)1(t ) ,..., NRC(S)Q( tt) : denote the nonresponse cells for the response type pair (St, ut).
(t )
Tot ijk
: denotes the number of units of type ut at site k of stratum j of PSU i, defined for each 1 ≤ t ≤ 3, 1
≤ i ≤ 16, 1 ≤ j ≤ 4, and each value of k for which site k∈RefSampij.
(t )
Resp ijk
: denote the number of units of type ut at site k of stratum j of PSU i for which we have a
response for St, defined for each 1 ≤ t ≤ 3, 1 ≤ i ≤ 16, 1 ≤ j ≤ 4, and each value of k for which site
k∈RefSampij.
Elig:={(i,j,k): site k in stratum j of PSU i is known to be eligible}
UnknownElig:={(i,j,k): the eligibility of site k in stratum j of PSU i is unknown}
es :=
∑
w
∑
wijk
ijk
(i, j,k )∈NRC (S )1s ∩Elig
for 1 ≤ s ≤ Q1
(i, j,k )∈NRC (S )1s \UnknownElig
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
55
∑
NRAdj (S )
(1)
s
+ es
w
ijk
(i, j ,k )∈NRC ( S )1s ∩ Elig
:=
∑
∑
w
ijk
(i, j ,k )∈ NRC ( S )1s ∩UnknownEli g
(1)
Resp ijk
wijk
for 1 ≤ s ≤ Q1
(i, j ,k )∈NRC ( S )1s ∩ Elig
sa(i,j,k): denotes the value v for which (i,j,k) ∈ NRC(S) (a)
v , defined for 1 ≤ a ≤ 3
∑
NRAdj (S ) (s2 ) :=
(2)
Tot ijk
NRAdj (S ) (1)
s1 (i , j ,k ) w ijk
(i , j ,k )∈NRC ( S ) 2s
∑
(2)
Resp ijk
NRAdj (S ) (1)
s1 (i , j ,k ) w ijk
for 1 ≤ s ≤ Q 2
(i , j ,k )∈NRC ( S ) 2s
2
∑
NRAdj(S)
(3)
s
∏ NRAdj(S)
(3)
Tot ijk
(i, j,k )∈NRC (S ) 3s
:=
wijk
for 1 ≤ s ≤ Q3
v=1
2
∑
Resp
(3)
ijk
(i, j,k )∈NRC (S ) 3s
′ := wijk
wijk
(v)
sv (i, j,k )
∏ NRAdj(S)
(v)
sv (i, j,k )
wijk
v=1
3
∏ NRAdj(S)
(v)
sv (i, j,k )
DurAdjijk
v=1
16
4
RefSamp := UU RefSampij
i =1 j =1
ObsChildijk (respectively, IntChildijk): denotes the number of children observed (respectively, have a
complete or partial interview) at site k of stratum j of PSU i, defined for all i,j,k for which k∈RefSampij.
w OBS :=
w INT :=
OBS
Trimijk
1
| RefSamp |
(i, j,k )∈RefSamp
1
| RefSamp |
(i, j,k )∈RefSamp
∑
ObsChild ijk
∑
′
IntChild ijk wijk
′
wijk
NRAdj(S) (3)
s3 (i, j,k )
⎧ 1, if wijk
′ ≤ 4.5 w OBS
⎪
:= ⎨ 4.5 w OBS
, otherwise
⎪ w′
ijk
⎩
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
56
TrimijkINT
⎧ 1, if wijk
′ ≤ 4.5 w INT
⎪
:= ⎨ 4.5 w INT
, otherwise
⎪ w′
ijk
⎩
C: denotes a characteristic of vehicle occupants
R: denotes a type of restraint
Bijk denotes the number of occupants of characteristic C restrained in restraint type R at site k of stratum j
in PSU i, defined for each (i,j,k) ∈ RefSamp.
Oijk denotes the number of occupants of characteristic C at site k of stratum j in PSU i, defined for each
(i,j,k) ∈ RefSamp.
∑
UseCR :=
3
TrimijkINT wijk ∏ NRAdj ( S )(sv)v (i, j,k ) DurAdjijk Bijk
(i, j,k )∈Re fSamp
∑
v =1
3
INT
ijk
Trim
(i, j ,k )∈Re fSamp
wijk ∏ NRAdj ( S )
v =1
, defined for all R and for those C
(v)
s v (i, j,k )
DurAdjijk Oijk
that involve at least one variable obtained by interview.
∑
UseCR :=
2
OBS
Trimijk
wijk ∏ NRAdj (S )(v)
s v (i, j,k ) DurAdjijk Bijk
(i, j,k )∈Re fSamp
∑
v =1
2
OBS
ijk
Trim
(i, j,k )∈Re fSamp
wijk ∏ NRAdj(S)
v =1
, defined for all R and for those C
(v)
s v (i, j,k )
DurAdjijk Oijk
that involve only variables obtained by observation.
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
57
13. References
Boyle, J., & Vanderwolf-Schulman, P. (2005, March). 2003 Motor Vehicle Occupant Safety
Survey, Volume 5: Child Safety Seat Report. Publication no. DOT HS 809 858. Washington, DC:
National Highway Traffic Safety Administration.
Cody, B. E., Mickalide, A. D., Paul, H. P., & Colella, J. M. (2002, February). Child Passengers at
Risk in America: A National Study of Restraint Use. Washington, DC: National SAFE KIDS
Campaign [now called Safe Kids Worldwide].
Reports, Forms, and Recordkeeping Requirements, Agency Information Collection Activity under
OMB Review, Federal Register Notice, Volume 71, Number 30, page 7824, February 14, 2006.
Revisions to the Standards for the Classification of Federal Data on Race and Ethnicity, Federal
Register Notice, Volume 62, Number 210, pages 58781-58790, October 30, 1997.
Glassbrenner, D. (2002, September). Safety Belt and Helmet Use in 2002 – Overall Results. DOT
HS 809 500. Washington, DC: National Highway Traffic Safety Administration.
Glassbrenner, D. (2005, February). Child Restraint Use in 2004 – Overall Results. DOT HS 809
845. Washington, DC: National Highway Traffic Safety Administration.
Glassbrenner, D. (To appear). The 2006 National Occupant Protection Use Survey –
Methodology Report, Washington, DC: National Highway Traffic Safety Administration.
Glassbrenner, D. (To appear). The Sample Redesign for the National Occupant Protection Use
Survey Washington, DC: National Highway Traffic Safety Administration.
Glassbrenner, D., & Ye, J. (2007, February). Child Restraint Use in 2006 – Overall Results,
National Highway Traffic Safety Administration, DOT HS 810 737. Washington, DC: National
Highway Traffic Safety Administration.
Glassbrenner, D., & Ye, J. (2007, August). Booster Seat Use in 2006. DOT HS 810 796.
Washington, DC: National Highway traffic Safety Administration.
Glassbrenner, D., & Ye, J. (2007, August). Child Restraint Use in 2006 – Demographic Results
DOT HS 810 797. Washington, DC: National Highway Traffic Safety Administration.
Glassbrenner, D., & Ye, J. (2007, August). Child Restraint Use in 2006 –Use of Correct Restraint
Types. DOT HS 810 798. Washington, DC: National Highway Traffic Safety Administration.
NHTSA. (2006, December). Motor Vehicle Traffic Crash Fatality Counts and Estimates of
People Injured for 2005. DOT HS 810 639. Washington, DC: National Highway Traffic Safety
Administration.
Office of Management and Budget, Update of Statistical Area Definitions and Guidance on Their
Uses, OMB BULLETIN NO. 05-02, February 2005.
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
58
Partners for Child Passenger Safety. (2004, July). The Forgotten Child. Philadelphia: The
Children’s Hospital of Philadelphia.
Partners for Child Passenger Safety. (2005, October). Fact and Trend Report. Philadelphia: The
Children’s Hospital of Philadelphia.
Public Law 106-414, 106th Congress, Transportation Recall Enhancement, Accountability, and
Documentation (TREAD) Act, 114 STAT. 1800, November 1, 2000.
Public Law 107-318, 107th Congress, Anton’s Law, 114 STAT. 1800, December 4, 2002.
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
59
14. Appendix
14.1 Data Collection Forms
The following forms were used by the data collectors to record the survey data. Each data collector was
given one “booklet” of forms for each site visit, plus additional booklets that they could use at a given site
if necessary (this will occur if they record data on more than 20 vehicles at a given site).
A “booklet” consisted of the form “Booster Seat Survey Recording Form” (displayed below, modified so
as to keep anonymous the names of the five fast food chains that participated in the survey) as a cover
page, followed by 20 copies of the untitled form displayed below following the Booster Seat Survey
recording form (i.e., the form that displays the survey’s OMB number).
The pages of the booklet were numbered. The displayed second form below, which has also been
modified to keep anonymous the participating fast food chains, shows Page 1.
Hi, My name is ____________ from Westat, a
national research organization. We are
conducting a Booster Seat Survey for the
National Highway Traffic Safety
Administration. We would simply like to
record the restraint use of everyone in your
vehicle and ask some simple questions. All
your responses and any observations I
make are completely confidential.
Can you tell me the Age of this child?
Height? Weight?
Survey Questions
1. In the past week, how many minutes did
this child spend in a vehicle driven by you?
Booklet #_____of ______
Booster Seat Survey Recording Form
Date: _ ____
PSU #: ________
Start Time for Booklet
_____
AM
PM
Site # : _
____
End Time for Booklet
_____
AM PM
2. In the past week how many times did you
visit a gas station when this child was in a
vehicle with you?
3. In the past week how many times did you
go to a (Chain 1) with this child in a vehicle
with you?
(Chain 2)? (Chain 3)? (Chain 4)? (Chain 5)?
4. In the past week how
to a recreation center
vehicle with y ou?
m any tim es did you go
when this c hild was in a
Which one(s )?
Site Type:
Gas Station Fast Food
Day Care Recreation Center
Site Name ______________________________________
Street Address _________________________________
Weather :
Light Precipitation
5. In the past week, how many times did you
go to a day care center with this child in a
vehicle with you?
Area :
Which one(s )?
Total Misses: ___
Urban
Total Refusals:___
NHTSA’s National Center for Statistics and Analysis
Suburban
Light Fog
Clear
Rural
Observer Name: _______________
1200 New Jersey Avenue SE., Washington, DC 20590
60
PSU #
Site #
MM / DD / YY
Car Pickup Van/SUV
Booklet # ____ of _____
1 st Row
2 nd Row
Question Key
Right
Driver Side
Middle
ADULT AGE
ADULT AGE
ADULT AGE
ADULT AGE
ADULT AGE
Restraint use
BELT
BELT
BELT
BELT
BELT
BELT
Race [ B W O ]
RACE
RACE
RACE
RACE
RACE
RACE
Gender [ M F ]
GENDER
GENDER
GENDER
GENDER
GENDER
GENDER
YY / MM
YY / MM
YY / MM
YY / MM
YY / MM
YY / MM
FT / IN
FT / IN
FT / IN
FT / IN
FT / IN
FT / IN
LBS.
LBS.
LBS.
LBS.
LBS.
LBS.
HH : MM
HH : MM
HH : MM
HH : MM
HH : MM
HH : MM
2. Gas station
GAS
GAS
GAS
GAS
GAS
GAS
(Chain 1)
McD
McD
McD
McD
McD
McD
(Chain 2)
Taco
Taco
Taco
Taco
Taco
Taco
(Chain 3)
BK
BK
BK
BK
BK
BK
(Chain 4)
Wendy
Wendy
Wendy
Wendy
Wendy
Wendy
(Chain 5)
KFC
KFC
KFC
KFC
KFC
KFC
#/
#/
#/
#/
#/
#/
NAME
NAME
NAME
NAME
NAME
NAME
Driver
Middle
Right
ADULT AGE
ADULT AGE
ADULT AGE
BELT
BELT
BELT
RACE
RACE
RACE
GENDER
GENDER
GENDER
YY / MM
YY / MM
FT / IN
FT / IN
Height
LBS.
LBS.
Weight
HH : MM
HH : MM
GAS
GAS
McD
McD
Taco
Taco
BK
BK
Wendy
Wendy
KFC
KFC
#/
#/
NAME
NAME
Teen (T)
Childre n O NLY(
(13 -15)
Young (Y)
(16 -24)
Adult (A)
(25 -69)
Senior (S)
#/
#/
NAME
NAME
Driver Side
Right
ADULT AGE
< 12 yrs)
Age
1. Time in Vehicle
3. Fast Food
Times in the Past Week:
OMB Permit No.: 2127 -0644
(70+)
Confidential Survey Information
WESTAT, Rockville, MD 20850
(301) 251-1500
Adult Age Group
Middle
Page 1
3 rd Row
4. Rec. Ctr. (Name)
5. Day Care (Name)
#/
#/
#/
#/
#/
#/
NAME
NAME
NAME
NAME
NAME
NAME
14.2 Letter of Authorization
The following letter of authorization was provided to the data collectors to show to people questioning
their authority to conduct the survey.
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
61
400 Seventh Street, S.W.
Washington, D.C. 20590
U.S. Department
of Transportation
National Highway
Traffic Safety
Administration
March 22, 2006
To Whom It May Concern:
Westat is under contract to the National Highway Traffic Safety Administration, U.S. Department
of Transportation, to conduct a survey on child transportation characteristics. The data collection phase
of the survey will take place from July 16, 2006, through August 1, 2006. It will consist of identifying
vehicles with child passengers and conducting brief interviews with those drivers at selected sites across
the country.
This county has been selected as one of the 16 areas across the United States that will be
surveyed. Information from this survey will be used to help design programs that improve the safety of
child passengers in motor vehicles.
Please direct any questions you may have to the Westat Project Director, [redacted], at [redacted].
Thank you very much for your support of this important research program.
Sincerely,
Donna Glassbrenner, Ph.D.
Program Manager
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
62
14.3 Incentives
The following incentives, which were stickers, were offered to potential respondents as incentives to
participate in the survey.
14.4 Frequently Asked Questions
The following list of questions and answers was provided to data collectors to help answer any questions
that potential respondents or others might have about the survey.
How was I selected for the survey?
As you drove on to the property I noticed that there were children in the vehicle.
How many people are you interviewing?
We are interviewing approximately 4,800 drivers of vehicles with child passengers throughout the
United States.
What is the purpose of this study? What is this survey about?
This survey will allow the government to compute national estimates of child safety seat use.
How long will this take?
The survey takes about 5 minutes to complete.
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
63
How will the study results be used? / What will you do with this information?
The survey will identify how children of different age groups are restrained when riding in a
vehicle.
How do I know you will keep this information confidential?
We are not collecting any personal identifying information. We will not be asking for names or
recording license plate information.
Why do you need to know the height and weight of my child(ren)?
The selection of appropriate child safety seats is dependent on age, height, and weight.
Why do you need to know how often my child goes to a gas station, fast food restaurant, daycare
center, or recreation center?
Department of Transportation will use this information to generate national estimates.
How will the results be published? Will the results be made available?
The Department of Transportation will publish the results of the study. However, that can take up
to a year or more. I can take your name and address and we can send you the study results
when they are available.
Do I have to do this? / Do I have to answer this survey? / I don’t want to do this.
You do not have to respond, but your help is very important to us. The information will provide
better national estimates on child safety seat use. You may refuse to answer any question at any
time.
Why don’t you ask someone else?
We are attempting to stop every adult who is driving with children in the vehicle to better
understand child safety seat use.
I had a bad experience recently with someone taking a survey, so I don’t think I want to
participate.
I’m sorry that your experience was unpleasant. We hope to make your contact with us a pleasant
and interesting experience. This is a legitimate research effort, in which your responses will help
us to learn about the use of child safety seats in passenger vehicles.
I think this whole business is stupid. The federal government could better spend my money. The
money for this study could be spent more wisely, etc.
[Occasionally you will encounter an argumentative respondent. In spite of their statements, this is
usually a person who is interested in the study, but wants to talk about what he feels before
consenting to complete the survey. Bear with him and hear him out! As long as he keeps talking,
he has not refused to do the survey. Do not argue: simply make short, neutral comments to let
him know you are listening.]
What is the authority/sponsor for this study?
The Department of Transportation is sponsoring this study. This survey is authorized by the
United States Code, Title 49, Section 111(c)(2).
Who do you work for?
I work for Westat, a survey research firm in the Washington, DC, area, and we have been
contracted by the Department of Transportation to conduct this study.
Who can I call at the Department of Transportation?
If you have questions about your rights as a person who is part of this study, please call the DOT
at: [redacted]. Please leave a short message with your name, phone number, and mention that
you are calling about the Booster Seat Survey. Someone will return your call as soon as
possible.
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
64
How do I know the survey is legitimate? / How do I know that you are really an interviewer for this
study?
If you would like, you can speak to my supervisor, or I can give you a toll-free number to call at
your convenience. The toll free number is: [redacted] and ask to speak with [redacted]. I also
have an authorization letter with the toll-free number on it.
Does this survey have approval from the Office of Management and Budget (OMB)? / What is the
OMB number?
Yes, the study has been approved by the Office of Management and Budget (OMB). The
approval number assigned to the study is 2127-0644 (it is listed on the side of the recording
forms).
14.5 Card for Race/Ethnicity Questions
The following information was provided to data collectors on a laminated card to show to interviewees to
aid in answering the race and ethnicity questions.
Are you of Hispanic or Latino origin?
1) Yes
2) No
What is your race? Please select one or more.
1) White
2) Black or African-American
3) Asian
4) Native Hawaiian or other Pacific Islander
5) American Indian or Alaska Native
National Survey of the Use of Booster Seats
OMB Control No. 2127-0644
NHTSA’s National Center for Statistics and Analysis
1200 New Jersey Avenue SE., Washington, DC 20590
65
DOT HS 811 111
April 2009
| File Type | application/pdf |
| File Title | Microsoft Word - 5832-NCSA Technical Report-April 2009.doc |
| Author | valeri.byrd |
| File Modified | 2011-03-10 |
| File Created | 2009-04-08 |