Estimation of Emigration from the United States using International Data Sources Jason P.

Download Report

Transcript Estimation of Emigration from the United States using International Data Sources Jason P.

Estimation of Emigration from the
United States using International
Data Sources
Jason P. Schachter
Senior Statistician,
Bureau of Statistics, ILO Geneva
United Nations Expert Group Meeting on Migration Statistics
New York, NY, December 4-7, 2006
Introduction




Many countries, including US, do not
collect emigration data.
Needed for the estimation of net
international migration.
Too problematic and expensive to collect
information on US citizens living abroad.
Attempt to use international data sources
to estimate number of US born/citizens
moving abroad.
2
How many US citizens live abroad?




State Department’s estimate was 4.1 million in
1999.
One-quarter (1 million) in Mexico, 700,000 in
Canada. Top 10 countries had 70% of all US
citizens living abroad
Huge discrepancies in totals when compared to
international data sources, though similar
country rankings.
State Department estimates appear to be two
to three times larger then international data
sources (closer to 1.5 to 2 million).
3
US State Department vs. International Data Sources
for Top 10 US Emigrant Countries
1,200,000
1,000,000
800,000
600,000
400,000
200,000
US State Dept
International data source
Sp
ai
n
nc
e
Fr
a
Au
st
ra
l ia
pin
es
Ph
i lip
Ita
ly
ae
l
Is
r
an
y
er
m
G
ng
do
m
Ki
Un
it e
d
Ca
na
da
M
ex
i co
0
4
About State Department data




State Department figures are “best guess” estimates
Voluntary registered + an estimated number of nonregistered
Purpose of data is in case of an emergency
evacuation, not to reflect exact numbers. Their
accuracy is of low priority to the State Department
Overestimation also possible due to failure to
deregister, as well as registration for short stays
abroad (compared to “usual residents” in other data
sources)
5
US military abroad






Inclusion or exclusion of US military personnel from data
sources adds to difficulties
According to pre-2001 military base report, one-quarter
million officially reported, plus an equal number of
dependents.
True size not reported for “security” reasons.
Does not include military stationed in Afghanistan and
Iraq, among other places.
Not normally included in foreign data (due to universe
coverage, residency rules), but US military, support staff,
and dependents living off-base could be included.
Needs to be kept in mind when estimating number of US
abroad
6
Using international stock data to estimate
“net migration”



Purpose of exercise was to investigate the
feasibility of using international data sources to
estimate the number of US citizens moving
abroad
Requested data on US citizens and/or US born
from the two most recent Censuses of 5
countries: Canada, France, Italy, Poland, and the
United Kingdom. (UNECE/EUROSTAT data
exchange initiative)
A number of measurement/data comparability
issues need to be kept in mind
7
US Citizens vs. US Born





By US law, US born is synonymous with US citizenship
Not treated as such in international data sources (selfreported)
For example, 2001 Spanish Census counted 21,000 US
born in US. 12,000 were foreigners, while 9,000 were
Spanish.
Dual “US-other” citizens usually not counted as US
citizens while living abroad (and only some countries Austria, Canada, Greece, Ireland, Japan, Malta, Portugal,
Switzerland, and most of Eastern Europe –collect this
information on their Censuses)
Conversely, limiting definition to US born misses US
citizens born abroad of American parents or naturalized
US citizens
8
Other measurement issues –data
comparability





Comparability between different data sources (Census,
registers, other administrative records, border
crossings).
Limit analysis to Census “stock” data to reduce data
comparability issues. No “flow” data.
Data quality --sampling and measurement error-- Did not
evaluate quality of data sources.
Different universes included in data sets, e.g. usual
residence definitions (de jure vs. de facto), how treat
those living temporarily abroad (e.g. students)
Need consistent universes between data sources (over
time within country, and between countries)
9
Country results





Requested the age and sex distribution of enumerated
US citizens and/or US born from the two most recent
Censuses.
France, Italy, and Canada asked both country of birth
and citizenship on their two most recent censuses
UK only asked country of birth
Poland collected this information on its 2002 Census, but
not on its 1988 Census, so not able to estimate
Only Canada and Poland collect information on dual
citizens (though not able to release Canada’s results due
to confidentiality restrictions)
10
US Born and US Citizens enumerated by 2001
Italian Census, by Sex
50,000
40,000
30,000
20,000
10,000
0
US Born
Female
Male
US
Citizens
Female
Male
11
Estimation uses a “crude” residual
methodology (in brief)


Observe stock data at two points in time (T1
and T2) from similar sources (e.g. Census)
Survive T1 population to T2 (using age and sex
specific death rates), for which there is a
comparable observed population. Difference
between the survived and observed population
is “net migration.”
12
Estimation Methodology (in more detail)




From two most recent Censuses, establish a T1 (circa
1990 Census) population of US born and/or citizens, and
a similar T2 (circa 2000 Census) observed population.
US born: T1 population is survived (using age and sex
specific death rates based on the resident US population
for the T1 year) a number of years equal to the
difference between T1 and T2.
The difference between the survived T1 and observed T2
population is assumed to measure “net international
migration” between the US and that given country over
the T2-T1 time period.
This figure is then divided by the time period to yield an
average annual net migration figure
13
Estimation Methodology (cont.)




Used same methodology for US citizens, but need two
additional components: birth of children to US female
citizens (added to the survived T1 total) and US citizens
who naturalized to another country –and revoked their
US citizenship-- (subtracted from the survived T1 total).
Proved difficult to incorporate additional components.
The number of births to US females is not necessarily
equal to a number of new US citizens (also misses
potential births to non-US nationals married to US
males).
Not able to find any public records on number of US
citizens who naturalized in these countries.
14
Method makes some dangerous
assumptions




The US population living abroad has the same age and
sex-specific death and birth rates as the US resident
population
Needed to make assumptions about the age
distribution of Americans living abroad (people were
equally distributed within each age cohort)
Improved coverage (measurement) of
foreigners/foreign-born from one Census to next
might be the actual reason for apparent “net
migration gain”
Measurement error around figures, from sample based
data, could be resulting in “net migration.”
15
(An even cruder) estimate of flows to and from
specific countries





Calculated a rough estimate of the in-flow of US
born/citizens to the US from specific countries, using the
US Census 2000 residence 5 years ago question, and
dividing by five.
Then used our net migration estimate combined with the
in-flow estimate to calculate an out-flow estimate.
Problems: US Census could include different universes than
other countries (e.g. students and military)
A five-year flow divided by five is not equivalent to a oneyear flow (migrants could have returned at any time during
the 5 year period, migrants could have left and returned
between 1996 and 2000, or they could have moved to
multiple countries during the 5 year period), all of which
underestimates the true size of country-specific flows.
Despite these problems, it is our best guess estimate of
flows to and from these countries.
16
Estimated Annual Inflow, Outflow, and Net
Migration of US Born to/from the United States
20,000
15,000
10,000
5,000
0
Canada
Great
Britain
Estimated Inflow
Estimated Outflow
France
Italy
Average annual net migration gain from US 17
Evaluation




Net migration results looked OK (at face validity,
given there are no sources to compare).
Canada and Great Britain had roughly the same
size of flows and net, despite a larger stock
population in Canada?
Large number of dual-nationals in Poland
(30,000 vs. 1,000 US citizens).
US born easier to calculate than US citizens
(dual citizen problem, components needed to
survive population) ---however citizenship data
are more readily available from different sources
18
Conclusions






How to deal with US military and dependents
Data quality issues—need to evaluate international
sources
Is this method possible combining different data sources
(Census, house hold surveys, population registers, etc.)?
Reliability of method is questionable, since it could mask
fluctuations over time.
Recommend using US born data to calculate
Recommend expanding project to include top thirty
receiving countries –but special tabulations from these
countries will be needed-
19
Contact Information
Jason Schachter, Ph.D
Bureau of Statistics, Room 5-51
4, Route des Morillons
CH-1211 Geneva, Switzerland
Phone: +41 (0)22 799 6954
E-mail: [email protected]
20