The 2006 American Community Survey (ACS) Content Test Questions on International Migration:...
-
Upload
carmel-hensley -
Category
Documents
-
view
217 -
download
0
Transcript of The 2006 American Community Survey (ACS) Content Test Questions on International Migration:...
The 2006 American Community Survey (ACS) Content Test Questions on
International Migration:
Improving Data on the U.S. Foreign-Born
Dean H. Judson
For presentation at the Conference of European Statisticians, Edinburgh, Scotland, 20-22 November, 2006
2
CES Recommendations for 2010 Censuses: Migration Section
“To facilitate and improve the
comparability of data at a regional level
through the selection of a core set of
census topics and the harmonization of
definitions and classifications.”
3
Issues with U.S. Census Bureau International Migration Data
• Difficult to obtain accurate distributions of the foreign born by U.S. citizenship status (citizenversus non-citizen)
• Difficult to estimate time spent in the United States by the foreign born
4
Opportunities for Improvement: The American Community Survey
• An ongoing nationwide survey that collects socioeconomic and housing information and
• replaces the long form componentof the 2010 census
5
Opportunities for Improvement: The ACS Content Test
• 63,000 housing units
• Two versions of question content 1) a control version and 2) a variant, or test, version3) Followup tested consistency of responses
• Changes that met data quality criteria will be implemented in the 2008 ACS, and reflected in the 2009 data release.
6
Improving Data on International Migration: Questions on the ACS
• U.S. Citizenship Status• Year of Entry
Related but Not Discussed Here. . .
• Place of Birth• Residence One Year Ago
7
Motivation for Changes: Citizenship Status• Naturalization: appears to be over-reported in some Census & survey data (Passel and Clark, 1997)
• Year of Naturalization: -item would help to reduce reports of naturalization by non-citizens (by prompting them to examine their answer)
-Year naturalized could be compared with year first entered to determine if respondent had beenin country long enough to naturalize
8
Motivation for Changes: Year of Entry
• Current question does not account for multiple entries (Redstone and Massey, 2004)
• Will better approximate host country experience by asking for first & most recent entry
9
Content Test Questions: Citizenship
Control Variant
10
Content Test Questions: Year of Entry
Control Variant
11
Summary of Question ChangesControl version:
Citizenship -five categories, including “Naturalized Citizen”
-American parent(s)
Year of Entry -Allows for reporting one entry
Test version:
Citizenship-five categories, including “Naturalized Citizen” and write-in for year naturalized
-U.S. Citizen Parents
Year of Entry -Allows for reporting more than one entry (first and most recent)
12
Selection Criteria: U.S. Citizenship Status
• The percent naturalized in the test version will be equal to or less than the percent for the control.
• The percent of non-responses in the test will be equal to or less than the percent for the control.
13
Citizenship Status: Naturalized Citizens
Universe: All nonblank records
5.1 5.2
0
5
10
15
20
25
Citizen by naturalization
Per
cent Test
Control
14
Citizenship Status Nonresponse Rates
1: Universe includes all nonblank records2: Universe includes all test records of naturalized citizens
9.8
2.7 3.00.0
5.0
10.0
15.0
20.0
25.0
Citizenship Status Year of Naturalization
Per
cent Test
Control
1 2
15
Summary: Citizenship Status
Percent Naturalized (Control versus Test)No statistical difference
Item Non-response:-Citizenship Status (Control versus Test) No statistical difference
-Year of NaturalizationTen percent non-response for those naturalized
Conclusion: Asking for year naturalized had no statistically significant effect on Citizenship item
but does have other uses
16
Selection Criteria: Year of Entry
• The net difference rate will be lower in the test version than the control (by period of entry).
• The percent of non-responses for the test version will be equal to or less than the percent for the control.
(for information purposes only. . .)• Determine if the year of entry values provided in the
control version reflect a first year of arrival, most recent year of arrival, or something else.
17
Number of Arrivals: Test versus Follow-up
39.4
60.6
87.2
12.80
20
40
60
80
100
Once More Than Once
Per
cent Test
Followup
47.7(4.6)
47.7(4.6)
Universe: Test cases of population born outside the U.S.
18
Year Entered: Control and Test versus Follow-Up
(Year entered matches exactly)
Universe: All persons born outside the United States.
68.074.0
66.1 65.770.3
63.5 63.5 60.2
0
20
40
60
80
100
2000 orlater
1990 to1999
1980 to1989
Before1980
Per
cent
Test (only or mostrecent arrival)
Control
19Universe: All persons born outside the United States.
80.590.0
84.1 87.783.7 84.1 84.7
94.1
0
20
40
60
80
100
2000 orlater
1990 to1999
1980 to1989
Before1980
Per
cent
Test (only or mostrecent arrival)
Control
Year Entered: Control and Test versus Follow-Up
(Year entered matches within two years)
20
Year of Entry Nonresponse RatesUniverse: Population born outside the United States
23.1 22.6
83.3
21.7
0
20
40
60
80
100
Year of First orOnly Arrival
Arrive More thanOnce?
Year of MostRecent Arrival
Per
cent Test
Control
1 21Year of First/Only Arrival
or Year of EntryArrive more than once?
Year of Most Recent Arrival
1: Includes all nonblank records.2: Includes all test cases that marked more than one entry.
21
Universe: All control records indicating more than one arrivalin reinterview.
Years Entered for Persons with Multiple Entries: Follow-up Interviews with Control Group to Check for
Consistency of Year Provided
18.0 18.8
32.6
44.6
25.1
21.3
9.1
30.6
0
5
10
15
20
25
30
35
40
45
50M
atc
hes B
oth
Matc
hes Y
ear
of F
irst A
rriv
al
Matc
hes Y
ear
of Last A
rriv
al
No M
atc
h
Matc
hes B
oth
Matc
hes
Decade/P
eriod
of F
irst A
rriv
al
Matc
hes
Decade/P
eriod
of Last A
rriv
al
No M
atc
h
Pe
rce
nt
Year to Year Same Decade
22
Summary: Year of Entry
Item Consistency
Number of Entries (Test versus Follow-Up):Large difference (48 percentage points) in proportion ofrespondents indicating single versus multiple entries.
Year Entered (Control and Test versus Follow-up):Somewhat consistent reporting of exact year and good reporting within two years.
Year Entered (Control vs. Follow-up, Multiple Entries):Follow-up reporting indicated that the original response more often represented the first year of arrival than the most recent year of arrival.However, exact year matches and same decade reporting was poor,with a sizeable proportion matching neither first nor last arrival.
23
Summary: Year of Entry (Cont.)
Item Non-responseYear Entered (Control versus Test):• Only or first arrival
-No statistical difference -non-response somewhat high for both versions (22 percent).
• Year of most recent arrival-very high non-response (83 percent)
Conclusion
The control version performed better, althoughfollow-up interviewing suggested concerns for thecontrol.
24
Thoughts on lessons learned • The purpose was to better represent the hard-to-
enumerate foreign born• Year of naturalization has analytic value
– Did no harm to the overall question– Can be used for consistency checking– Appears to be well understood
• Year of arrival– Despite successful cognitive testing…– Question form continues to be problematic– Many inconsistent responses