The Net Undercount of Children in the Decennial Census Based on

Report
The Net Undercount of Children in
the Decennial Census Based on
Demographic Analysis
by
Dr. William P. O’Hare
O’Hare Data and Demographic Services, LLC
COPAFS Presentation,
June 7, 2013
Why Focus on The Undercount of
Young Children in the Census?
• Young children have had high net
undercount rates historically in the
United States
• The net undercount rates of young
Children have been increasing since
1980
• There is very little systematic scientific
evidence about this problem
2
Presentation and Terminology
• “Undercount” versus “Difference”
• Net Undercount Here = Census - Estimates
• So negative number implies an undercount
• Positive number implies an overcount
3
How Do We Know Who Is Missed
In The Census?
• Demographic Analysis (DA)
Compares census results to an
independent estimate based largely
on birth and death certificate data
• Dual-Systems Estimates (DSE)
Compares census results to a
second follow-up survey conducted
in selected areas (Called Census
Coverage Measurement in 2010)
4
Percent Difference Between Census Counts and DA and
DSE Estimates for Young Children: 2000 and 2010
5.0
4.6
4.0
Percent Difference
3.4
3.0
2.6
2.0
1.0
0.7
0.2
0.0
-1.0
-0.5
2000 (age 0-9)
2010 (age 0-9)
DA
2010 (age 0-4)
DSE
Source: O'Hare et al. 2012 SDA Presentation
5
What is Demographic Analysis?
• Estimates for population under age 75 are
based on historical components of change
for cohorts:
[births, deaths, net international migration]
P = B – D + NIM
• 99.6% of age 0-4 DA estimate is based on
births
6
Components of DA Estimates for
Age 0-4
• Births = 21,076,000
• Deaths = 148,000
• Net International Migration = 244,000
Source: Census Bureau’s May 2012 DA Release
7
Limitations of DA Estimates
• Only National Level Data
• Only Black and Non-Black Data
Historically
• Only Net Undercount/Overcount Figures
• No Estimation of Uncertainty
8
Overall Results of
2010 Decennial
Census Look Good
9
Comparison of Various 2010
Population Figures (in millions)
310
308.5
308.3
308.7
308
306
304
302
300.7
300.7
CCM Pop.
Census
CCM Count
300
298
296
2010 Pop. DA Middle
Est
Series
Census
Count
10
Good Overall
Results Mask
Important
Differentials
11
Percent Difference Between 2010 Census Counts and DA
Estimates for Children (age 0-17) and Adults (age 18+)
1
0.7
0.5
P
e
r
c
e
n
t
D
i
0
f
f
e -0.5
r
e -1
n
c
e -1.5
-2
0.1
Total
Children
-1.7
Adults
Source: Velkoff 2011, PAA Presentation
12
Percent Difference Between 2010 Census Counts and
DA Estimates by Single Year of Age: 0-17
3.0
P
e
r
c
e
n
t
D
i
f
f
e
r
e
n
c
e
2.0
1.0
0.0
0
1
2
3
4
5
6
7
8
9
10 11 12 13 14 15 16 17
-1.0
-2.0
-3.0
-4.0
-5.0
-6.0
AGE
Source: Census Bureau’s May 2012 DA Release
13
Percent Difference Between 2010 Census Counts and DA
Estimates for Race and Hispanic Groups by Single Year of Age:
0-17
8.0
P
e
r
c
e
n
t
D
i
f
f
e
r
e
n
c
e
Hispanic
6.0
4.0
Black Alone
2.0
0.0
1
2
3
4
5
6
7
8
9
10
-2.0
-4.0
12
13
14
15
16
17
18
Black Alone or in
Combination
Not Black Alone
or in
Combination and
Not Hispanic
-6.0
-8.0
-10.0
11
AGE
Note: Data on Hispanics is only available from the December 2010 DA Release.
The “Not Black Alone or in Combination and Not Hispanic” category is not a category used by the Census Bureau.
The racial categories used here are the Modified Race Categories where people who marked “some other race” were
assigned to one of the five major races categories,
14
Difference Between 2010 Census and DA for People Under
Age 5, by Sex, Race and Hispanic Origin
TOTAL
FEMALE
MALE
BLACK ALONE OR IN
COMBINATION
HISPANIC
NOT BLACK ALONE OR IN
COMBINATION AND NOT
HISPANIC
Number
Percent
972,000
471,000
501,000
-4.6
-4.5
-4.6
247,000
-6.3
414,000
-7.5
309,000
-2.6
Note: Data on Hispanics is only available from the December 2012 DA Release.
The “Not Black Alone or in Combination and Not Hispanic” category is not a category used by the Census Bureau.
The racial categories used here are the Modified Race Categories where people who marked “some other race” in the Census
are assigned to one of the major race categories.
15
How Does the Data form
2010 Compare to Earlier
Censuses?
16
Mean Percent Difference Between Census Counts and DA
Estimates by Single Year of Age from 1950 to 2010
P
e
r
c
e
n
t
D
i
f
f
e
r
e
n
c
e
0.0
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
-0.5
-1.0
-1.5
-2.0
-2.5
-3.0
-3.5
-4.0
-4.5
Age
Source; Census Bureau's May 2012 DA Release and Internal Census Bureau Historic file
17
Percent Difference Between Census Counts
and DA Estimates for Adults and Young
Children: 1950 to 2010
1.0
P
e
r
c
e
n
t
D
i
f
f
e
r
e
n
c
e
0.0
Ages 0-4
1950 1960 1970 1980 1990 2000 2010
-1.0
-2.0
Adults age
18+
-3.0
-4.0
-5.0
Source: Census Bureau’s May 2012 DA Release and internal Census Bureau Historic File
18
Key Findings
•
Children have a net undercount & adults have a net overcount in 2010 Census
•
Net undercount rates of children vary by age and race/ethnicity
•
Younger children (under age 5) have the highest net undercount rate (4.6%) in
2010 and there is a net overcount for persons age 14 to 17
•
Black and Hispanic children account of most of high net undercount of young
children and net overcount of 14-to-17-year-olds.
•
High net undercount of young children is not new…..The age structure of net
undercounts for children is relatively consistent since 1950
•
1950 to 1980, net undercount rates for adults and young children fell, but 1980
to 2010, net undercount of adults fell while undercount of young children
increased
19
KEY QUESTIONS
1. Why are there such high net undercount rates for
young children in the census?
2. Why is there a net overcount rate for 14-17 yearolds in the 2010 Census?
3. What accounts for the strong correlation between
net undercount rates and age among children?
4. Why has the net undercount rate for young
children increased since 1980?
20
Some Potential Reasons for High
Net Undercount of Young Children
1. DA Estimates for young children are too high
2. Problems in collection and processing of
data in Census
3. Time constraints among parents of young
children
4. Young children live in households and
families that are difficult to enumerate
21
Hypothesis - DA Estimates
for young children are too
high
Emigration of young children
• Pitkin and Parks (2005) hypothesize many
children born to foreign-born (Mexican)
women move to Mexico at a young age
and are not picked up in DA emigration
statistics.
• But, 0-4 year-olds missing in 2000 were
found as 10-14 year-olds in 2010.
23
Hypothesis - Systematic
errors in census-taking or
processing lead to net
undercount of young children
The Census Questionnaire
• Continuation Form …Persons 7-12 need
follow up (NRFU Persons 6-12)
• Young children are likely to be listed last
on Census Questionnaire
• 10% of young children live in 7+ person
households compare to 3% of adults
25
Several Improvements in the 2010
Census Form
• Added age to information collection on
persons 6-12 on primary questionnaire
• Added a new “administrative” question
about people who were left off roster
• Added instruction about “including babies”
• Added instruction about “child custody”
• Partnership with American Academy of
Family Physicians and Planned
Parenthood
26
Despite improvements, net
undercount of young children
increased between 2000 and
2010
Age Imputation
Hypothesis - Too many people who
should have had age imputed as 04 got age imputed as 14-17 (or
other age groups)
28
Age Allocation Rates
4.0
3.5
P
e
r
c
e
n
t
A
l
l
o
c
a
t
e
d
3.0
2000
Census
2.5
2.0
2010
ACS
1.5
1.0
0.5
0.0
0
2
4
6
8
AGE
10
12
14
16
Source: Analysis of PUMs
files on IPUMS site
29
Hypothesis - Parents of
young children don’t
complete the census
questionnaire because of
time demands
30
2000 Census Mail-Back Rates by Presence of
Children and Family Type
Census Mail-Back
Rate
Live Alone
86%
Single with adult roommates
73%
Single with kid(s)
63%
Married with no kids
90%
Married with kid(s)
83%
Source: Hillygus, Nie, Prewitt & Pals, 2006, Table
4.4
31
Hypothesis - Young children live in
the kinds of households and living
arrangements that are difficult to
enumerate
Living Arrangements of Young
Children
• 10 characteristics of Hard-to-Count
populations identified by Census Bureau in
Planning Data Base
• Young (age 0-4) Black and Hispanic children
higher on every one of the 10
33
80
Percent of Adults, Black Children Age 0-4 and
Hispanic Children Age 0-4 with Selected Hard-toCount Characteristics
70
70
60
60
P
e
r
c
e
n
t
Adults
50
46
Black age
0-4
43
40
37
36
35
Hispanic
age 0-4
31
30
23
21
20
16
15
10
0
In Building with
2+ units
In Poverty
In Rental unit
Moved Last Year
Source: Analysis of 2010 ACS PUMS file on IPUMS system
Summary
• Young Children have higher net undercount rate than
any other age group in 2010
• Young children have had relatively high net
undercount rates since 1950
• The trends in net undercount rates of young children
and adults have diverged since 1980
• Need to develop understanding of WHY young
children have high net undercount rates in the
census
• One focus of 2020 Census planning should be
households with young black or Hispanic children
35
THANKS
Contact Information
William O’Hare
[email protected]
36

similar documents