Date post: | 25-Dec-2015 |
Category: |
Documents |
Upload: | lester-ray |
View: | 214 times |
Download: | 0 times |
Improving the estimation of long-term international emigration at local authority level
Joshua Turner
Population Statistics Research Unit (PSRU)
Local Insight Reference Panels
1
Session Topics
Brief Background
Work so far
Preliminary results: Impact of changes
Current work
Next steps
2
Brief Background
3
Importance of Emigration
• Birmingham
Source: Population Estimates for UK, England and Wales, Scotland and Northern Ireland, Mid-2011 and Mid-2012, ONS
Mid-2011 Population 1,074,283
Births +17,636
Deaths -8,028
International In-Migration +11,710
International Out-Migration -7,002
Movers to elsewhere in UK -45,503
Movers from elsewhere in UK +42,338
Mid-2012 Population 1,085,417
4
Brief Background
• No datasets with robust counts of emigration at Local Authority level
• Current method uses a model-based approach
5
Stepwise Model
Stepwise modelStepwise model
Relationship between
Relationship between
IPS LA Level
Emigration Estimates
(3 Year Average)
Predictor Variable 1Predictor Variable 1
Variable 2Variable 2
Variable iVariable i
Variable 3Variable 3
Variable 4Variable 4
Variable 5Variable 5
Predictor Variable 1Predictor Variable 1
Variable 2Variable 2
Variable 4Variable 4
Variable 3Variable 3
6
Predictor Variables
7
Poisson Regression Model
Emigration estimates
constrained to IPS totals
Poisson Regression
Model
IPS LA Level
Emigration Estimates (3
Year Average)
Predictor Variable 1Predictor Variable 1
Variable 2Variable 2
Variable 4Variable 4
Variable 3Variable 3
8
A B D C E
New Migration Geographies (NMGos)
9
REGION Z
NMGO 1 NMGO 2
Local Authority
NMGo
Region
Cluster Analysis
Constraining Emigration Estimates
REGION Z
NMGO 1 NMGO 2
Local Authority
NMGo
Region
6,00021,000Predicted Estimates
Final Estimates
20,000 5,000
25,000 Emigrants
10
Work Carried Out So Far
11
Investigate a Non-Modelling
Approach
Update the Current Method
• Explored a non-modelling approach • Greater use of administrative data sources• Closely similar to the immigration method
o BUT unlike the immigration method, there are no datasets which directly count emigration at Local Authority level
The Non-Modelling Approach
12
Non-Modelling Approach: Streaming Migrants
IPS England and Wales National Emigration Estimate
Reason for Migration
Study
E.g.Higher Education Statistics Agency (HESA) Student Record
Work
E.g. Lifetime Labour Market Database (L2)
Other
Children 17-59 60 +
E.g.Patient Register Data System (PRDS)
13
• Lifetime Labour Market Database (L2)o 1% sample of records on the National Insurance
and Pay as You Earn System (NPS)o Economic activityo 12 months or more of economic inactivity as an
indicator of possible emigration
• IPS ‘Other’ Categoryo Out-migrants coded as ‘Other’ when free-text
answer related to ‘Work’ or ‘Study’ reason
Non-Modelling Approach: Data sources
14
• Promising results, however...o Improvements in data sources neededo Timing considerations need more development
• Research and results will help inform how we update the current emigration model
Source: Emigration user update August 2014, ONS
Non-Modelling Approach: On pause
15
Updating the Current Model
• Part 1: Removal of NMGos• Part 2: Preliminary Results• Part 3: Investigating Predictor Variables
16
Remove the Intermediate Geography – New Migration Geography (NMGo)
Local Authority population size used in the model to account for area differences (as a rate)
Constraining to the IPS-based region level estimates
Intermediate (NMGo)
Local Authority
Regional
Part 1: Removing NMGos
17
Reviewed:• Research Review Group (ONS Panel of
Experts)• Consultation with University of Southampton
Approved:Removing NMGosPoisson regression methodUsing LA population size to account for area
differences
Part 1: Removing NMGos
18
Part 2: Preliminary Results
Removing the NMGos
19
Without NMGo Model & Current Model
20
Birmingham
Wandsworth
Newham
Tower Hamlets
Camden
Manchester
EalingBrent
Haringey
Lambeth
Lewisham
Hackney
Southwark
Richmond upon Thames
Kensington & Chelsea
Westminster
Hammersmith & Fulham
City of London
Cardiff
Oxford
Leeds
Birmingham
Wandsworth
Newham
Tower Hamlets
Camden
Manchester
Ealing
Brent
Haringey
LambethLewisham
Hackney
Southwark
Richmond upon Thames
Kensington & Chelsea
Westminster
Hammersmith & Fulham
City of London
Cardiff
Oxford
Leeds
IPS LA Outflows & Current Model
2121
City of Bristol
Stafford
Birmingham
Wandsworth
Newham
Tower Hamlets
Camden
Manchester
Ealing
Brent
Haringey
Lambeth
Lewisham
Hackney
Southwark
Richmond upon Thames
Kensington & Chelsea
Westminster
Hammersmith & Fulham
City of London
Cardiff
Oxford
Leeds
IPS LA Outflows & Without NMGo Model
22
City of Bristol
Stafford
IPS Outflow – London Region (2012): 102,683
Case Study: London NMGos
Model NMGo Predicted Constrained
Current Model LOI1 26,502 25,351
LOI2 26,623 25,510
LOI3 24,253 23,133
LOI4 17,134 16,404
LOI5 12,940 12,286
Without NMGo Model
LOI1 26,213 25,047
LOI2 23,164 22,134
LOI3 20,296 19,396
LOI4 27,367 26,153
LOI5 10,414 9,952
23
NMGo Case Study: LOI3 (Predicted)
24
NMGo Case Study: LOI3 (Final)
25
NMGo Case Study: LOI4 (Predicted)
26
NMGo Case Study: LOI4 (Final)
27
Part 3: Investigating Predictor Variables
28
Project is currently researching potential predictor variables:
Less reliance on Census data Greater use of administrative data sources More intuitive
Using ‘manual selection’ of predictor variables
Data sources investigated thoroughly at LA level
Part 3: Investigating Predictor Variables
29
University of Southampton consultations approve:
Stepwise and manual selection of predictor variables
But ensure no correlation between predictor variables and LA population size
Part 3: Investigating Predictor Variables
30
Data sources which are being explored include:• Patient Register Database• HESA Student Record• HESA Destination of Leavers Survey• Lifetime Labour Market Database (L2)• Migrant Worker Scan• English School Census• Welsh School Census• Annual Population Survey• Home Office Crime Statistics
Part 3: Investigating Predictor Variables
31
• Population aged 64 and over • Students (aged 20 to 25) of Non-UK nationality in their final year
of study• Students (aged 20 to 25) of Non-EU nationality in their final year
of study• Students of Non-EU nationality in their final year of study• In-migrants of EU8 nationality registering for a National
Insurance number• Employed individuals, aged 16 and over• Long-term international in-migration flows• Short-term international in-migration flows • Household with accommodation owned outright• Household with accommodation owned with a mortgage• Higher/further education students of non-UK nationality
Part 3: Possible Predictor Variables
32
Next Steps
1) Continuing research into updating the method
2) Assessment and comparisons
3) Review of changes by RRG and University of Southampton
4) Further consultations with users
33
Key Points
Non-modelling approach on pause Emigration method updated:
1) Removing NMGos
2) Updating Predictor Variables
Impacts of updates are being investigated
34
What other local data sources should we be exploring?
What are your thoughts on the ‘manual selection’ of predictor variables?
35