The MIMOSA Model for Estimating International Migration in the European Union

1
The MIMOSA Model for Estimating International
Migration in the European Union
James Raymer and Guy Abel February 2008
MIgration MOdelling for Statistical
Analyses (Eurostat Project 2006/S 100-106607/EN,
LOT 2)
2
MIMOSA project

• Aim of project
• Development and application of statistical
  modelling techniques for the estimation of
  missing data on migration flows and foreign
  population stocks
• Work with National Statistical Agencies to
  improve their data
• Funded by Eurostat, December 2006-2009
• Netherlands Interdisciplinary Demographic
  Institute (Coordinator)
• Central European Forum for Migration and
  Population Research
• Southampton Statistical Sciences Research
  Institute
• GéDAP, Université Catholique de Louvain
• Focus of this presentation Migration flow
  estimation

3
Significance

• The study of international migration in the
  European Union is currently hindered by data
  availability, quality and consistency
• Harmonization of data collection processes and
  the data they generate is not even close to being
  realized, therefore our understanding of
  population change and migration policy is
  therefore currently limited
• To overcome these obstacles, we need models to
  (i) harmonise and correct for inadequacies in the
  available data and (ii) estimate the missing
  patterns

4
Data used in this study

• Provided by Eurostat (i.e., data comes from the
  Joint Questionnaire on International Migration)
• Immigration and emigration flows by previous /
  next country of residence (32 by 32 matrix)
• 31 European countries EU plus Iceland,
  Liechtenstein, Norway and Switzerland
• Rest of world
• Years 2002, 2003, 2004 and 2005

5
Double-entry matrix for selected countries, 2003
I Receiving countrys reported flow E
sending countrys reported flow no reported
data available
6
Double-entry matrix for selected countries, 2003
I Receiving countrys reported flow E
sending countrys reported flow no reported
data available
7
Double-entry matrix for selected countries, 2003
I Receiving countrys reported flow E
sending countrys reported flow no reported
data available
8
Data
Immigration Immigration Emigration Emigration
Timing Other EEA Nationals Other EEA Nationals
Instant DE, IE, IT, LU, ES BE, EE, FI, DE, IE, IT, LV, LU, ES EE, DE, IE, LU, ES BE, EE, DE, IE, LU, ES
3 months AT, BE, EE, HU, SI AT, HU, SI AT, BE, HU, SI AT, HU, SI
6 months DK, IS, LT, NL, NO DK, IS, LT, NL, NO DK, LV, NO DK, LV, LT, NO
12 months CY, CZ, FI, LV, PT, SE, CH, UK CY, SE, CH, UK CY, IT, FI, LT, NL, PT, SE, CH, UK CY, IT, FI, NL, PT, SE, CH, UK
Permanent MT, PL, RO, SK CZ, MT, PL, RO, SK CZ, MT, PL, RO, SK CZ, MT, PL, RO, SK
Unknown BG, FR, GR, LI BG, FR, GR, LI, PT BG, FR, GR, IS, LI BG, FR, GR, IS, LI

• Only partial data exists between 2001 and 2004
• Italics No data exists between 2001 and 2004

9
Methodology

• Harmonise and clean available data
• Identify base data (DK, FI, SE and NO)
• Calculate adjustment ratios and adjust reported
  data by using iterative procedure
• Estimate missing data
• Immigration and emigration totals for European
  matrix
• Associations between origins and destinations

10
Available data by receiving countries
11
Available data by sending countries
12
Step 1 of iterative adjustment procedure
Identify and adjust base data
Note, procedure applied to 2002-2005 data pooled
over time
13
Step 2 of iterative adjustment procedure
Calculate adjustment ratios
Note, procedure applied to 2002-2005 data pooled
over time
14
Step 3 of iterative adjustment procedure Add
adjusted flows to base data
Note, procedure applied to 2002-2005 data pooled
over time
15
Adjustment factors for receiving country (I) and
sending country (E) migration data, 2002-2005
16
Final adjusted matrix
17
Estimated coefficients for regressions on natural
logarithms of available immigration and
emigration flows from / to EU matrix and rest of
world (n 59)
18
Multiplicative component model
Origin by destination migration flow table
Multiplicative components where
19
Estimating the ODij assocations

• ODij reported to expected flows
• Expected flows obtained using iterative
  proportional fitting (IPF) and estimated margins
  from EU matrix
• Two sets of ODij
• Available data
• Missing data

20
Estimated coefficients for regression on
available origin-destination associations within
the EU matrix (n2540)
21
Estimated migration totals, 2002-2005 within the
EU matrix and from / to rest of world
22
Top 20 estimated immigration and emigration
countries (in thousands), 2002-2005
Immigration Emigration
23
Estimated emigration flows from Poland, 2002-2005
24
Estimated emigration flows from France, 2002-2005
25
Summary

• To produce an overall picture of EU migration,
  one has to address issues concerning the
  availability, quality and consistency of
  migration data
• Harmonisation and cleaning of reported data
• Estimation of missing data (marginal totals and
  origin-destination associations) based on
  patterns found in available data and covariates

26
Next steps

• Much research still needed
• Integration
• EM and Bayesian methods
• Standard errors of estimates
• MIMOSA project
• Flows by age, sex and citizenship
• Flows to be linked to population estimates and
  change