Appendix A

2017 National Survey of Children’s Health sample
frame
Keith Finlay
Center for Administrative Records
Research and Applications
US Census Bureau
keith.ferguson.finlay@census.gov
301-763-6056
March 2, 2017
This document describes using administrative records to build a sample frame for the National
Survey of Children’s Health (NSCH).

Population of interest
The population of interest is all children residing in housing units in the US on the date of the
survey.

A sample frame for all households with children
The sample frame identifies three mutually exclusive strata:
• [1] Households with explicit links to children in administrative data.
• [2a] Households without explicit links to children in administrative data, but predicted to be
likely to have children conditional on administrative data.
• [2b] Households without explicit links to children in administrative data, but predicted to be
unlikely to have children conditional on administrative data.
This document first explains the construction of the Stratum 1 flag, then documents the separation of
Strata 2a and 2b.

1

Stratum 1: identifying explicit links from children to addresses
The Stratum 1 flag for all households with explicit links to children comes from three data sources:
the Numident, a list of Social Security Number applicants with data updated from various administrative records; and the CARRA kidlink file, a prototype linkage between children and parents
based on Census and administrative records. Household addresses are updated with the Master
Address Auxiliary Reference File, a file that links person identifiers with the latest location updates
from a variety of administrative data.

Using the Numident to identify children
The Numident is based on all individuals who have been assigned Social Security Numbers.
Demographic data from the Numident is updated from federal tax data and various administrative
records. There are 75,156,219 children in the December 2016 Numident who will be aged 0–17
years on April 1, 2017 . Figure 1 shows the distribution of date of birth for these children.

0

5.0e−05

Density
1.0e−04

1.5e−04

2.0e−04

Figure 1: Distribution of date of birth, aged 0–17 years as of April 1, 2016, December 2015
Numident

02apr1998

01jan2002

01jan2008
Date of birth

01jan2014

Identifying the households containing children in the Numident
To sample households with children, we must connect the children in the Numident to the households
in which they live. We do this with the CARRA kidlink file.
CARRA kidlink
The CARRA kidlink file uses data from Census survey and federal administrative records to link
children Protected Identification Keys (PIKs) to parent PIKs. We can use this file to identify the
parents of children in the Numident.
2017 National Survey of Children’s Health sample frame

2

The source data for the CARRA kidlink file are: the Census Numident, the 2010 Census Unedited
File, the IRS 1040 and 1099 files, the Medicare Enrollment Database (MEDB), Indian Health
Service database (IHS), Selective Service System (SSS), and Public and Indian Housing (PIC) and
Tenant Rental Assistance Certification System (TRACS) data from the Department of Housing and
Urban Development. Of these, the IRS 1040 provides the most significant information.
In the CARRA kidlink file generated March 2016, there are 75,156,219 unique records for
children who will be aged 0–17 years on April 1, 2017 .
Let us consider how many children from the Numident have been linked to a parent in the
CARRA kidlink file. Table 1 shows the number of children linked with both a mother and a father,
linked with a mother only, linked with a father only, or not linked with any parent.
Table 1: Child-parent links in the CARRA kidlink file relative to the Numident population, aged
0–17 years as of April 1, 2016, March 2016 CARRA kidlink file
Type of link
Frequency Percent
Mother and father
50,886,028
68%
Mother only
14,643,347
19%
Father only
3,050,257
4.1%
No link
6,576,587
8.8%
All children in Numident 75,156,219
100%
Figure 2 compares the distributions of date of birth for these children against the distribution
shown in Figure 1.

0

Frequency
5.0e+05
1.0e+06

1.5e+06

Figure 2: Frequency distributions of date of birth, Numident vs. kidlink entries, aged 0–17 years as
of April 1, 2016

02apr1998

01jan2002

01jan2008
Date of birth
Numident

01jan2014
kidlink entries

The CARRA kidlink file will be updated in March 2017 for NSCH sample frame production.

2017 National Survey of Children’s Health sample frame

3

Updating household location using the MAF-ARF
In order to update household location, we use a Census dataset called the Master Address Auxiliary
Reference File (MAF-ARF). The MAF-ARF links person identifiers to address identifiers using
Census survey data and federal administrative data. The source data for the MAF-ARF file are:
the Census Numident, the 2010 Census Unedited File, the IRS 1040 and 1099 files, the Medicare
Enrollment Database (MEDB), Indian Health Service database (IHS), Selective Service System
(SSS), and Public and Indian Housing (PIC) and Tenant Rental Assistance Certification System
(TRACS) data from the Department of Housing and Urban Development, and National Change
of Address data from the US Postal Service. Of these, the IRS 1040 provides the most significant
information.
Out of 75,156,219 children in the Numident, 59,841,686 are matched directly to a MAFID. Out of
65,529,375 kidlink-matched mothers, 60,031,595 are matched to a MAFID. Out of 53,936,285
kidlink-matched fathers, 49,767,120 are matched to a MAFID.
For each child observation from the Numident, we now have four possible MAFIDs: the SSI
MAFID, the kid to MAF-ARF MAFID, the child-to-kidlink-to-mother-to-MAF-ARF MAFID,
and the child-to-kidlink-to-father-to-MAF-ARF MAFID. I allocate the single MAFID using that
order. First, I assign the SSI MAFID (1,265,823 cases). If MAFID is missing, I assign the directly
identified child MAFID (58,805,360 cases). If the MAFID is still missing, I assign the mother
MAFID (6,385,866 cases). Finally, if the MAFID is still missing, I assign the father MAFID
(2,101,661 cases). That leaves 6,597,509 children from the Numident not assigned MAFIDs (a
MAFID match rate of 87.2%).
There are some MAFIDs associated with a great number of children. As an example, out of
68,558,710 children associated with a MAFID, 296,808 children are associated with a MAFID with
more than 20 child-MAFID links.
The 68,558,710 children associated with a MAFID are then collapsed down to 36,642,194
unique MAFIDS. This implies 1.87 children per household for households assigned a flag.
We then need to scale up the MAFID list to the universe of MAFIDs to allow sampling of
unflagged households. A merge of the 36,642,194 unique child-flagged MAFIDS with the January
2017 ACS MAF-X file matches 36,609,700 MAFIDS with child flags, removes 32,494 MAFIDS
with child flags, and adds 159,897,403 MAFIDs without child flags. The sample frame file now has
196,507,103 valid MAFIDS, of which 36,609,700 MAFIDS include child flags . Compare this with
the 2011 ACS, in which 37,147,503 out of 114,991,725 households included related children.1
The MAF-ARF will be updated in March 2017 for NSCH sample frame production.

Stratum 1 construction visualization
Figure 3 shows a visualization of the sample frame construction.
1

http://www.census.gov/prod/2013pubs/p20-570.pdf

2017 National Survey of Children’s Health sample frame

4

Figure 3: Sample frame construction
Numident: children
in population
SSI recipients:
kids to MAFIDs

MAF-ARF: kids
to MAFIDs

Kidlink: kids
to moms and dads

MAF-ARF: moms
to MAFIDs

MAF-ARF: dads
to MAFIDs

Set of PIK-MAFID links
(four possible)

Set of child-flagged MAFIDs
(collapse from PIK-level to MAFID-level;
prioritizing SSI, then child MAFID,
then mother MAFID, then father MAFID)
Complete set of ACS household MAFIDs
(append valid MAFIDs,
exclude invalid MAFIDs)

Strata 2a and 2b: identifying probabilistic links from children to
addresses
In 2016, the Stratum 1 flag performed well. That is, it contained approximately the same rate of
children after sampling, as had been predicted before the survey. The survey team would like to
further increase the sampling efficiency of the survey by adding more information to the second
stratum. By definition, Stratum 2 does not have explicit links from children to households in the
administrative data. In 2017, we will further bifurcate Stratum 2 into those households more likely
to have children and those households less likely to have children.
Households will be assigned to Stratum 2a based on a model of child presence as a function of
variables available in administrative data for all households in the MAF. The model is estimated
with data from the most recent year of the ACS, in which child presence can be observed. Then
parameter estimates from that model can be used to predict the likelihood of child presence for all
households. These models are estimated separately for each state, and the threshold for bifurcation
is based on an objective of maximizing the size of Stratum 2b while also maintaining 95% coverage
of households with children in Strata 1 and 2a.

2017 National Survey of Children’s Health sample frame

5

Definitions
Population or sample concepts
• 2015 ACS sample, edited and swapped
– unit of observation is the household, unless noted otherwise
– sample includes sampled vacant dwellings, unless noted otherwise
• MAF
– population but restricted to MAFIDs marked as valid for ACS
Sample frame notation
•
•
•
•

h indexes household
s indexes states
C equals 1 if a household has any children, 0 otherwise
Strata:
– S1 : household with children
– S2a : household likely to have children
– S2b : household unlikely to have children

• Strata sizes:
– p(S1 )
– p(S2a )
– p(S2b )
• Strata child rates:
– p(C|S1 )
– p(C|S2a )
– p(C|S2b )
• Coverage with unsampled S2b :
– p(S1 ∪ S2a |C)

Model
Our goal is a scalar measure of the likelihood of a child being associated with a MAFID. This
measure must be available for all ACS-valid MAFIDs in the MAF. Using a sample in which the
presence of children is observable, we will estimate a model of child presence. The regressors used
to make the index prediction must be observable for all MAFIDs (i.e., to predict outside of the
estimation sample to the entire MAF).

2017 National Survey of Children’s Health sample frame

6

The general model is:
Ch = f (Xh ; θ),
where C is equal to one if a household includes any children and zero otherwise, X is a vector of
characteristics available for all households, and θ is an unknown vector of parameters.
We estimate the model using the most recent ACS 1-year sample:
E[Ch |Xh ] = f (Xh ; βˆACS ) for households h in the ACS.
With parameter estimates from the ACS, we make predictions for the entire MAF:
Cˆh = f (Xh ; βˆACS ) for households h in the MAF.
In practice, we estimate models separately for each state. We do this to account for systematic
differences in administrative records coverage and MAF quality across states. The model can now
be specified as:
E[Chs |Xhs ] = f (Xhs ; βˆs,ACS ) for households h in state s in the ACS,
where s is the MAFID’s state and the parameters βˆs,ACS now vary across states. The state-specific
predictions become:
Cˆhs = f (Xhs ; βˆs,ACS ) for households h in state s in the MAF.

Estimation
The model above is estimated as a linear probability model separately for each state using the edited
and swapped 2015 ACS sample. The outcome is child_present, a flag for whether a child is
present at the sampled MAFID.
The following covariates are included (with associated data sources) and are available for each
MAFID (except where a missingness flag is used):
• 2015 ACS 5-year published aggregate data
– acs_blkgrp_childrate_lvout: proportion of residents of block group who are
children, excluding the own-observation child counts from the numerator and denominator
• MAF-ARF
– female2050: flag for female between ages 20 and 50 at MAFID
– adult2050: flag for adults between ages 20 and 50 at MAFID
– coresid_sexdiff: flag for coresidence of men and women between ages 20 and
50 at MAFID
– miss_adult2050: flag for missingness from MAF-ARF
2017 National Survey of Children’s Health sample frame

7

• IRS 1040 filings, tax year 2015
– any_kid_deduct_max: does any tax form associated with this MAFID have any
deduction related to children?2
– itemized_max: does any tax form associated with this MAFID use itemized deductions?
– miss_any_kid_deduct_max: flag for MAFIDs without associated tax forms
• VSGI NAR commercial data
– vsgi_nar_homeowner_max: does any observation associated with this MAFID
record it as homeowener-occupied?
– miss_vsgi_nar_homeowner_max: flag for MAFIDs without associated VSGI
data
• Targus commercial data
–
–
–
–
–
–
–
–

targus_homeowner_0: various flags for homeowner-occupied MAFID
targus_homeowner_A: various flags for homeowner-occupied MAFID
targus_homeowner_B: various flags for homeowner-occupied MAFID
targus_homeowner_C: various flags for homeowner-occupied MAFID
targus_homeowner_D: various flags for homeowner-occupied MAFID
targus_homeowner_E: various flags for homeowner-occupied MAFID
targus_homeowner_F: various flags for homeowner-occupied MAFID
miss_targus_homeowner: flag for MAFIDs without associated Targus data

Parameter estimates are stored in the file frame2017_child_present_bystate.csv.

Sample frame objective function
In order to choose an optimal Strata 2a, we use the following objective function:
• Minimize the size of Strata 2a while maintaining coverage of at least 95%
Strata 2a is defined as:
S2a = {households in the MAF with Cˆh > C¯ but not in S1 }.
Strata 2b is defined as
S2b = {households in the MAF but not in S1 or S2a }.
With state-specific modeling, the objective function and coverage constraint also becomes state
specific:
2

The following IRS variable were used to make this variable: child exemptions and EITC qualifying children.

2017 National Survey of Children’s Health sample frame

8

• Minimize the size of Strata 2a in each state while maintaining coverage of at least 95% in
each state
State-specific Strata 2a is defined as:
S2a = {households in the MAF with Cˆhs > C¯s but not in S1 }.
Strata 2b is defined as before.

Optimization algorithm
The optimization parameter is a threshold on the child-present prediction probability, such that
MAFIDs with values above the threshold are assigned to Stratum 2a. Starting at a low threshold
¯ 3 , follow this algorithm:
(C)
¯ calculate the proportion of MAFIDs in Stratum 2a, p(S2a ),
1. Under the current threshold C,
and the coverage of Strata 1 and 2a under no sampling of Strata 2b, (p(S1 ∪ S2a |C)).
2. If p(S2a ) > 0 and p(S1 ∪ S2a |C) ≥ 0.95, then increase the child prediction threshold C¯ one
step (e.g., 0.01) and return to (1). If p(S1 ∪ S2a |C) < 0.95, then the previous threshold C¯ is
the optimal cutoff for S2a .
Under state-specific modeling, this algorithm is applied separately to each state.

Optimal strata
Table 2 shows the optimal strata under a 95% coverage constraint for Strata 1 and 2a. The coverage
constraint assumes non-sampling of Stratum 2b. The notation is as defined above. The strata were
optimized separately for each state using parameter estimates from separate state regressions of
child presence in the 2015 ACS microdata.

Auditing the sample frame against the ACS
To examine the performance of the administrative records used to build the sample frame, we merge
the list of MAFIDs constructed above with the American Community Survey housing-unit sample
from 2014. Currently, this audit uses unedited ACS data (i.e., item nonresponse are left as missing
and are not imputed including children’s age). If item nonresponse is random with respect to the
presence of children in the household, this should not cause any systematic bias in the audit.
All estimates are weighted with the housing-unit-level weights, which include weight for vacant
units. In vacant housing units, we assign zero children. These estimates should reflect the NSCH
survey production process.
3

The most conservative starting threshold would be at p(S1 ), where p(S2b ) = 0.

2017 National Survey of Children’s Health sample frame

9

Table 2: Optimal 2017 NSCH strata with 95% coverage constraint, state-level optimization
State
N
p(S1) p(S2) p(S3) p(C|S1) p(C|S2) p(C|S3) p(C|!S1) p(!S3|C) q CˆS2
US 2,305,707
0.24
0.33
0.43
0.78
0.17
0.03
0.10
0.95 43 0.10
AL
37,368
0.22
0.38
0.40
0.74
0.16
0.03
0.10
0.95 38 0.14
AK
9,394
0.19
0.62
0.19
0.71
0.19
0.05
0.15
0.95 20 0.13
AZ
44,652
0.22
0.34
0.43
0.76
0.20
0.03
0.10
0.95 43 0.14
AR
22,498
0.22
0.43
0.36
0.77
0.17
0.04
0.11
0.95 34 0.14
CA 215,206
0.28
0.31
0.41
0.82
0.21
0.04
0.11
0.95 43 0.14
CO
37,863
0.24
0.33
0.44
0.84
0.18
0.03
0.10
0.95 43 0.14
CT
23,154
0.24
0.30
0.46
0.80
0.17
0.03
0.09
0.95 43 0.14
DE
7,156
0.22
0.19
0.59
0.78
0.18
0.02
0.07
0.96 46 0.16
DC
4,697
0.20
0.37
0.43
0.71
0.12
0.02
0.07
0.95 33 0.09
FL
120,642
0.21
0.29
0.50
0.72
0.15
0.02
0.08
0.95 50 0.13
GA
56,483
0.27
0.34
0.40
0.76
0.19
0.04
0.11
0.95 40 0.15
HI
9,819
0.17
0.54
0.28
0.72
0.24
0.04
0.17
0.95 28 0.13
ID
11,439
0.24
0.32
0.44
0.77
0.18
0.03
0.09
0.95 38 0.16
IL
97,181
0.25
0.33
0.42
0.79
0.18
0.03
0.10
0.95 43 0.14
IN
47,914
0.25
0.27
0.48
0.78
0.19
0.03
0.09
0.95 47 0.17
IA
33,959
0.23
0.24
0.53
0.83
0.16
0.03
0.07
0.95 50 0.17
KS
26,796
0.25
0.27
0.48
0.79
0.17
0.03
0.08
0.95 46 0.17
KY
33,621
0.23
0.36
0.41
0.77
0.17
0.03
0.10
0.95 40 0.14
LA
30,576
0.25
0.37
0.38
0.70
0.16
0.04
0.10
0.95 37 0.15
ME
17,240
0.15
0.45
0.40
0.79
0.10
0.02
0.07
0.95 36 0.12
MD
38,955
0.27
0.28
0.45
0.80
0.18
0.03
0.09
0.95 44 0.15
MA
42,884
0.24
0.30
0.47
0.82
0.16
0.03
0.08
0.95 46 0.14
MI
100,300
0.22
0.25
0.53
0.80
0.16
0.02
0.07
0.95 54 0.16
MN
72,731
0.23
0.21
0.56
0.84
0.18
0.03
0.07
0.95 56 0.17
MS
18,393
0.25
0.42
0.33
0.71
0.17
0.04
0.12
0.95 31 0.14
MO
50,265
0.23
0.31
0.46
0.79
0.18
0.03
0.09
0.95 46 0.16
MT
11,319
0.17
0.47
0.36
0.76
0.15
0.03
0.09
0.95 33 0.13
NE
20,792
0.24
0.27
0.49
0.82
0.17
0.03
0.08
0.95 46 0.17
NV
18,322
0.24
0.33
0.43
0.74
0.17
0.03
0.10
0.95 40 0.13
NH
11,118
0.19
0.31
0.50
0.81
0.13
0.02
0.07
0.95 41 0.14
NJ
56,618
0.25
0.31
0.43
0.82
0.19
0.04
0.11
0.95 43 0.14
NM
16,057
0.18
0.47
0.35
0.73
0.18
0.03
0.11
0.95 33 0.14
NY 137,665
0.23
0.40
0.38
0.76
0.18
0.03
0.11
0.95 38 0.12
NC
68,734
0.23
0.35
0.42
0.78
0.15
0.03
0.09
0.95 41 0.14
ND
9,543
0.20
0.36
0.44
0.80
0.16
0.03
0.08
0.95 38 0.15
OH
89,658
0.24
0.28
0.48
0.80
0.16
0.03
0.08
0.95 49 0.16
OK
45,705
0.23
0.45
0.32
0.73
0.19
0.04
0.12
0.95 33 0.15
OR
26,908
0.23
0.30
0.47
0.82
0.16
0.03
0.08
0.95 44 0.14
PA
118,520
0.22
0.30
0.48
0.81
0.15
0.03
0.08
0.95 49 0.14
RI
6,630
0.22
0.34
0.44
0.79
0.15
0.03
0.09
0.95 34 0.12
SC
32,766
0.23
0.32
0.45
0.74
0.15
0.03
0.09
0.95 43 0.14
SD
9,981
0.22
0.32
0.45
0.79
0.19
0.03
0.09
0.95 40 0.17
TN
43,918
0.25
0.32
0.43
0.77
0.16
0.03
0.09
0.95 43 0.15
TX 146,469
0.28
0.34
0.38
0.78
0.21
0.04
0.13
0.95 40 0.15
UT
18,497
0.33
0.28
0.39
0.83
0.23
0.04
0.12
0.95 37 0.19
VT
8,924
0.15
0.46
0.39
0.83
0.12
0.03
0.08
0.95 33 0.13
VA
53,822
0.26
0.28
0.46
0.81
0.16
0.03
0.09
0.95 46 0.15
WA
47,760
0.25
0.30
0.45
0.82
0.19
0.03
0.10
0.95 44 0.14
WV
15,110
0.17
0.49
0.34
0.74
0.18
0.03
0.12
0.95 31 0.12
WI
75,272
0.22
0.23
0.56
0.83
0.17
0.02
0.07
0.95 56 0.17
WY
4,413
0.20
0.46
0.34
0.74
0.17
0.03
0.11
0.96 28 0.13

2017 National Survey of Children’s Health sample frame

10

Table 3 shows the overlap between the MAFID and ACS distributions with respect to whether
any children were present in the household.
Table 3: Comparison of NSCH child flags and ACS data, any children in household, 2014 ACS,
housing unit weights including vacants
Observed ACS households
NSCH child flags
No children Any children
Total
No children
92.2%
7.8%
100.0%
Any children
25.2%
74.8%
100.0%
Total
74.6%
25.4%
100.0%
N (ACS households) 2,322,722

Child flag performance by age group
We are particularly interested in the coverage of young children. In this section, we show how the
child flags perform for specific age groups. These are stricter tests since any deviation in age beyond
the age interval will cause either a Type 1 or Type 2 error.
Table 4 shows the overlap between the MAFID and ACS distributions with respect to whether
any children aged 0–2 years were present in the household. Given that the input administrative
records used to construction the child flags are 1–2 years old and that the ACS data are from 2014,
it is not surprising that the overlap for children aged 0–2 years is much lower than the overall rate
shown in Table 3.
Table 4: Comparison of NSCH child flags and ACS data, any children in household aged 0–2 years,
2014 ACS, housing unit weights including vacants
Observed ACS households
NSCH child flags
No children 0–2 Any children 0–2
Total
No children 0–2
99.0%
1.0%
100.0%
Any children 0–2
71.8%
28.2%
100.0%
Total
97.5%
2.5%
100.0%
N (ACS households)
2,322,722
Table 5 shows the overlap between the MAFID and ACS distributions with respect to whether
any children aged 3–5 years were present in the household. By ages 3–5, overlap between the child
flag and the ACS data is above 60%.
Table 6 shows the overlap between the MAFID and ACS distributions with respect to whether
any children aged 6–8 years were present in the household.
Table 7 shows the overlap between the MAFID and ACS distributions with respect to whether
any children aged 9–11 years were present in the household.
Table 8 shows the overlap between the MAFID and ACS distributions with respect to whether
any children aged 12–14 years were present in the household.
2017 National Survey of Children’s Health sample frame

11

Table 5: Comparison of NSCH child flags and ACS data, any children in household aged 3–5 years,
2014 ACS, housing unit weights including vacants
Observed ACS households
NSCH child flags
No children 3–5 Any children 3–5
Total
No children 3–5
97.1%
2.9%
100.0%
Any children 3–5
38.5%
61.5%
100.0%
Total
92.9%
7.1%
100.0%
N (ACS households)
2,322,722

Table 6: Comparison of NSCH child flags and ACS data, any children in household aged 6–8 years,
2014 ACS, housing unit weights including vacants
Observed ACS households
NSCH child flags
No children 6–8 Any children 6–8
Total
No children 6–8
96.9%
3.1%
100.0%
Any children 6–8
35.7%
64.3%
100.0%
Total
92.3%
7.7%
100.0%
N (ACS households)
2,322,722

Table 7: Comparison of NSCH child flags and ACS data, any children in household aged 9–11
years, 2014 ACS, housing unit weights including vacants
Observed ACS households
NSCH child flags
No children 9–11 Any children 9–11
Total
No children 9–11
96.9%
3.1%
100.0%
Any children 9–11
33.4%
66.6%
100.0%
Total
92.1%
7.9%
100.0%
N (ACS households)
2,322,722

Table 8: Comparison of NSCH child flags and ACS data, any children in household aged 12–14
years, 2014 ACS, housing unit weights including vacants
Observed ACS households
NSCH child flags
No children 12–14 Any children 12–14
Total
No children 12–14
97.0%
3.0%
100.0%
Any children 12–14
32.3%
67.7%
100.0%
Total
92.1%
7.9%
100.0%
N (ACS households)
2,322,722

2017 National Survey of Children’s Health sample frame

12

Table 9 shows the overlap between the MAFID and ACS distributions with respect to whether
any children aged 15–17 years were present in the household.
Table 9: Comparison of NSCH child flags and ACS data, any children in household aged 15–17
years, 2014 ACS, housing unit weights including vacants
Observed ACS households
NSCH child flags
No children 15–17 Any children 15–17
Total
No children 15–17
96.9%
3.1%
100.0%
Any children 15–17
31.3%
68.7%
100.0%
Total
91.9%
8.1%
100.0%
N (ACS households)
2,322,722

State-specific performance
Table 10 shows the overlap between the MAFID and ACS distributions by state. The smallest
oversample strata are in Hawaii, Maine, Vermont, and West Virginia. The largest oversample strata
are in California, Texas, and Utah. The highest rates of Type 1 error are in DC, Florida, Louisiana,
Mississippi, Nevada, and South Carolina. The highest rates of Type 2 error are in Alaska, Hawaii,
New Mexico, Texas, and Utah.

Small-area paper-only response probability
Since 2012, the American Community Survey (ACS) respondents have been able to submit survey
forms over the internet in addition to completing and mailing back a paper questionnaire. We
used 2016 ACS response mode choices summarized at the block group and other block group and
tract-level characteristics to model Web and paper response mode probabilities by block group.
Sample households will be located within block groups and assigned a paper-only response
probability. The 30% of households with the highest paper-only response probability will be
flagged as ‘High Paper’ and will receive a paper questionnaire with the initial web invitation. For
very new housing units without assigned Census blocks, we assign a value of zero for this binary
variable (i.e., the default for these new households is high Internet accessibility.)
In addition to ACS response mode, we modeled block group response mode probabilities using
multinomial logisitic regression on adult education (%HS, %College, %Grad), poverty (%in
poverty, %between 100% and 150% of poverty threshold), adult age, race, ethnicity, foreign
born, and rural/urban status. Modeled response mode probabilities are given a household weight
of 10 and averaged with obserserved ACS response mode probabilities to offset sampling error
in very small samples.

Local-area household income relative to the poverty rate
The frame has a set of poverty variables from the 2015 5-year American Community Survey file.
These variables measure the proportion of households with household income in an interval defined
by the poverty rate. Figure 5 shows the kernel-smoothed probability distribution function of the
proportion of households in the block group that have household income less than 150% of the
poverty rate.
2017 National Survey of Children’s Health sample frame

13

Table 10: Comparison of NSCH child flags and ACS data, any children in household, 2014 ACS,
housing unit weights including vacants, by state
NSCH frame
Any children (a)
No children (b)
N (c)
ACS obs. children Any (d) None (e) Total Any (f) None (g)
Total
(d)/(a)
(e)/(a)
(a)/(c) (f)/(b)
(g)/(b) (b)/(c)
(c)
State
×100
×100
×100 ×100
×100
×100
Alabama
71.5
28.5
24.3
7.5
92.5
75.7 37,511
Alaska
72.7
27.3
23.7
12.6
87.4
76.3
9,534
Arizona
73.1
26.9
23.6
8.5
91.5
76.4 44,646
Arkansas
70.4
29.6
25.0
8.7
91.3
75.0 22,495
California
78.3
21.7
29.7
9.3
90.7
70.3 217,111
Colorado
80.2
19.8
25.3
8.2
91.8
74.7 37,691
Connecticut
77.6
22.4
24.9
7.3
92.7
75.1 23,385
Delaware
73.0
27.0
25.1
6.4
93.6
74.9
7,367
District of Columbia
64.9
35.1
19.3
5.7
94.3
80.7
4,693
Florida
66.7
33.3
22.7
6.1
93.9
77.3 121,828
Georgia
71.6
28.4
29.9
8.9
91.1
70.1 57,019
Hawaii
71.3
28.7
18.0
16.5
83.5
82.0
9,856
Idaho
76.4
23.6
27.2
8.0
92.0
72.8 11,545
Illinois
75.0
25.0
26.5
8.1
91.9
73.5 97,583
Indiana
75.6
24.4
27.2
7.0
93.0
72.8 48,569
Iowa
79.2
20.8
26.5
5.8
94.2
73.5 34,025
Kansas
76.6
23.4
27.7
7.2
92.8
72.3 26,961
Kentucky
73.1
26.9
25.7
8.5
91.5
74.3 34,115
Louisiana
66.6
33.4
27.8
8.9
91.1
72.2 31,206
Maine
75.3
24.7
17.7
5.1
94.9
82.3 17,636
Maryland
76.3
23.7
28.5
6.8
93.2
71.5 39,331
Massachusetts
78.3
21.7
24.8
6.8
93.2
75.2 43,395
Michigan
76.9
23.1
24.2
5.4
94.6
75.8 100,990
Minnesota
81.0
19.0
25.9
5.2
94.8
74.1 72,611
Mississippi
69.3
30.7
27.8
9.3
90.7
72.2 18,761
Missouri
74.7
25.3
24.7
6.8
93.2
75.3 50,595
Montana
74.4
25.6
19.0
7.6
92.4
81.0 11,567
Nebraska
80.4
19.6
27.4
6.4
93.6
72.6 21,002
Nevada
69.3
30.7
25.6
8.5
91.5
74.4 18,288
New Hampshire
78.0
22.0
21.8
5.9
94.1
78.2 11,239
New Jersey
79.2
20.8
26.3
8.3
91.7
73.7 57,087
New Mexico
71.2
28.8
22.0
10.4
89.6
78.0 16,173
New York
72.4
27.6
23.9
9.1
90.9
76.1 138,735
North Carolina
73.7
26.3
25.7
7.4
92.6
74.3 68,857
North Dakota
76.2
23.8
25.1
6.5
93.5
74.9
9,642
Ohio
75.5
24.5
25.9
6.2
93.8
74.1 90,191
Oklahoma
70.4
29.6
26.6
9.7
90.3
73.4 46,397
Oregon
79.0
21.0
23.6
7.0
93.0
76.4 26,748
Pennsylvania
76.6
23.4
23.9
5.8
94.2
76.1 120,084
Rhode Island
75.9
24.1
22.8
7.7
92.3
77.2
6,819
South Carolina
69.3
30.7
25.1
6.8
93.2
74.9 32,989
South Dakota
76.4
23.6
23.6
7.3
92.7
76.4
9,957
Tennessee
73.2
26.8
26.6
7.7
92.3
73.4 44,043
Texas
73.8
26.2
31.6
10.7
89.3
68.4 146,897
Utah
80.1
19.9
35.6
10.5
89.5
64.4 18,761
Vermont
79.5
20.5
17.6
7.0
93.0
82.4
9,097
Virginia
76.6
23.4
28.3
7.1
92.9
71.7 54,668
Washington
76.3
23.7
26.0
7.5
92.5
74.0 47,839
West Virginia
70.5
29.5
18.3
9.1
90.9
81.7 15,434
Wisconsin
79.7
20.3
24.1
5.8
94.2
75.9 75,291
Wyoming
72.5
27.5
22.6
9.8
90.2
77.4
4,458
2017 National Survey of Children’s Health sample frame

14

Figure 4: Kernel-smoothed probability distribution function of anticipated response mode
probabilities

0

1

Density

2

3

Figure 5: Kernel-smoothed probability distribution function of block-group-level 150% poverty
rate, ACS, 2015 5-year file

0

.2
.4
.6
.8
1
Proportion of individuals below 150% of poverty line, weighted, by block group

2017 National Survey of Children’s Health sample frame

15

Final sample frame data layout
The component data files are merged together based on MAFID. The data layout for this combined
file is given in Table 11.
Table 11: NSCH population data file layout
Level of
Variable name
Label
variation
mafid
Master Address File ID
MAFID
maf_curstate
State
State
maf_curcounty
County
County
maf_curblktract
Tract
Tract
maf_curblkgrp
Block group
Block group
maf_curblk
Block
Block
stratum1
Stratum 1 identifier
MAFID
stratum2a
Stratum 2a identifier
MAFID
stratum2b
Stratum 2b identifier
MAFID
acs_tract_net_response ACS Internet response
Tract
blkgrp_lt_100_povrate Pr. HH w/ inc. < 100% poverty rate Block group
blkgrp_100_150_povratePr. HH w/ inc. 100–150% poverty rateBlock group
blkgrp_150_185_povratePr. HH w/ inc. 150–185% poverty rateBlock group
blkgrp_185_200_povratePr. HH w/ inc. 185–200% poverty rateBlock group
blkgrp_gt_200_povrate Pr. HH w/ inc. > 200% poverty rate Block group
blkgrp_lt_150_povrate Pr. HH w/ inc. < 150% poverty rate Block group
Filename: nsch_pop_file.sas7bdat
Population: all MAFIDs in January 2017 MAF-X
Unit of observation: household (MAFID)
Number of observations: 196,507,103
Filesize: 26GB

2017 National Survey of Children’s Health sample frame

Type
long
str2
str3
str6
str1
str4
byte
byte
byte
float
float
float
float
float
float
float

Domain
9 digits

{0, 1}
{0, 1}
{0, 1}
[0, 1]
[0, 1]
[0, 1]
[0, 1]
[0, 1]
[0, 1]
[0, 1]

Any
missing?
no
no
no
yes
yes
yes
no
no
no
yes
yes
yes
yes
yes
yes
yes

16