improved method for the geographical distribution of out-migrants
DESCRIPTION
Improved Method for the Geographical Distribution of Out-Migrants. Fiona Aitchison and Jonathan Swan. Former Method. Previous year’s resident population used to apportion HA/FHSAs. New Method. Geographic Level. Data/Methods. (Including new visitor switcher assumptions). - PowerPoint PPT PresentationTRANSCRIPT
Improved Method for the Geographical Distribution of
Out-Migrants
Fiona Aitchison and Jonathan Swan
Former Method
England and Wales
GOR/Wales
Local Authorities
Intermediate Geography: HA/FHSAs
National
Published IPS
3 Year IPS average used to apportion GOR
North East
Newcastle &
North Tyneside
Newcastle upon Tyne North Tyneside
Geographic Level Data/Methods
Previous year’s resident population used to apportion
HA/FHSAs
New Method
National
GOR/Wales
LAs
Propensity to Migrate model
used to apportion NMGo
3 Year IPS average used to apportion GOR
Published IPS
New Migration Geography for emigrants (NMGo)
Newcastle upon Tyne North Tyneside 5 Other LAs
NEI1
North East
England and Wales
Geographic Level Data/Methods
(Including new visitor switcher assumptions)
Factors available in the model
• Armed Forces• Crime• Education• Employment• Ethnicity• Housing• Deprivation• Migration• Existing population• Socio-economic classification• Students• Tenure• Country of Birth
Modelling methods considered
• Factor Analysis followed by Enter method Linear Regression
– Created 4 or 5 components built from approximately 20 of the available 100+ variables.
– Model gave an R2 value of approximately 68%– Disadvantage: Complex with hard to interpret results
• Forward-Stepwise Linear Regression– Created model with 3 variables selected from the
available 100+ – Model gave an R2 value of approximately 78%
Modelling methods considered
• Forward-Stepwise Linear Regression with logged variables
– Model gave an R2 value of approximately 75%– Disadvantage: A number of variables could not have
logarithm taken
• Forward-Stepwise Linear Regression (direct count of out-migrants)
Testing procedure
• Precision of model measured using the Average Square Error (ASE) on a number of test sets of data
• Log model was found to be subject to bias towards underestimation
Modelling Method Indicative ASE
Factor Analysis 0.45
Stepwise Regression (Propensity) 0.26
Stepwise Regression (Logs) 0.28
• The stepwise regression model of propensity to migrate was selected due to more plausible results
Example of the model: 2006
• In 2006 the variables below are used to form the model, in addition to a constant term.
• Estimated in-migrants• Males aged 16-34 with limiting long-term illness• Persons in higher professional occupations• Females aged 40-44• Percentage of males in population
• Model results in a significant improvement – The percentage of variance explained is increased– R2 increases from around 40% to over 80%– In 2006 R2 is 91%
Changes from Indicative Results
• Indicative results for revised 2002 to 2005 estimates were published in April 2007
• An additional variable, Country of Birth, was included in the list of factors
• The intermediate geography was revised for the West Midlands and Wales
• The models for these years have all changed slightly in terms of the variables selected
Future Work
• It is not intended to change the modelling methodology for at least the next two years
– The model will still be updated each year with new data– Results from extra out-migrant filter shifts on IPS will
become available
• Further research in this area will be taken forward as part of wider migration research
International Migration Sex Ratios
Sex Ratio – Methodology Considered
• Group LAs into quartiles and/or quintiles• In Migrants
– Grouped by sex ratio of Census one year ago resident outside UK
• Out Migrants– Grouped by sex ratio of resident LA population.– Groups fixed by 2001 ratios and– Variable groups by previous years population
considered.
• Research undertaken by Michelle Littlefield, ONSCD
Sex Ratio – example grouping Out migrants, quintiles, variable membership
0.6
0.8
1
1.2
1.4
1.6
2001 2002 2003 2004 2005
1
2
3
4
5
Quintile
Sex ratios – Out MigrantsLondon vs. Non London
0.6
0.8
1
1.2
1.4
1.6
2001 2002 2003 2004 2005
Out of London London
Sex ratios - Conclusions
• All the variants we examined for LA groupings produced broadly similar results.
• Therefore, not able to determine stable groupings of LAs for sex ratios.
• London / non-London split produced results we were not able to explain.
• Therefore unable to produce method for sex-ratios of international migrants.
• So the national sex-ratio is used.• Subject of possible further research.
International Out Migration Age Distribution
International Migration (IPS) age profiles
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0-4 5-9 10-14 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70+
Age
Pro
po
rtio
n o
f to
tal m
igra
nts
in a
ge
gro
up
non-British out-migrants
non British - in-migrants agedon 2 years
Allocation of British / Non British
0%
10%
20%
30%
40%
50%
60%
70%
Mid-02 Mid-03 Mid-04 Mid-05Mid-year
% n
on
-Bri
tish
IPS
co
nta
cts
1 2 3 4 5 2-4
Age Distribution of British out-migrantsGrouping LAs (Males)
0
1
2
3
4
5
6
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80
Age
Ou
t-m
igra
nts
%
Cluster 1 Cluster 2
Summary of Age Distribution Approach
• For each sex separately• Split into British non-British
– LAs grouped by quintiles – middle three grouped– Quintiles on in-migrants as % of resident population
• Non British– Use age individual LA distribution of in-migrants …– but aged on two years
• British– Split into two clusters– Clusters based on resident population age distribution– Use IPS quintile age distribution– Split to SYOA based on Census in-migrant distribution
• Research undertaken by Karen Gask
Any Questions…