MERA Monitoring and Evaluation Research Associates LtdPANZ99

1
James Newell - MERA

• First Steps Towards a Multi-Regional Migration
  Model for New Zealand

2
Approach

• Systems Development / End use perspective
• Discussion of selected issues in Developing the
  Knowledge Base / Underlying Migration Series

3
System Development Approach

• Set a long term goal / desired features
• Design and develop the simplest system as a first
  step toward that goal - don't try and build
  anything too complex in the first iteration of
  system development
• Practical considerations of time / resources
  determine size of the steps towards desired final
  design
• Modular design of system components then
  integrate these together (object oriented
  approach)
• Establish some means for assessing performance
  criteria of enhancements / new features added to
  the initial prototype
• From initial prototype - incrementally evolve the
  system towards the final goal, adapting and
  learning from the experience of each step

4
Migration Model Development

• End Use Requirements and Data availability
• Resulting Definitions / Modelling Framework
• Development of Raw Knowledge Base - Raw Migration
  Estimates and Derivatives
• Description of Migration Patterns
• Explanation of Migration Behaviour
• Projection of Future Migration Patterns
• Design needs to reflect
• - nature of the end use / application
• - data requirements for explanatory model
  development

5
Data Sources

• "Form of the Migration Model is defined as much
  by the nature of available data as it is by the
  purpose" (Stillwell and Congdon 1991)
• Census of Population and Dwellings (5 yrly) -
  subnational
• International Arrival and Departure Records
  (Monthly) - national and some subnational
• Data available offshore on NZ emigrants location,
  composition

6
Why This Problem

• Demographically based models for service and
  infrastructure planning at a Sub-National Level
• Improve Modelling of Future Development in the
  Wellington Region - currently on inferred net
  flows modelled
• Analysis of Patterns of Population Change in
  Rural NZ over 1986-96 at a Regional Level
• Contributing to Waikato Universities New
  Demographic Directions Programme

7
Migration Definitions

• Involves a change of residence by a person,
  family or household
• Can be short distance or long
• Can be short duration or long (temporary or
  permanent/long term)
• Residential Migration is a subset of all
  population movements for all purposes

8
Migration Definitions (Cont.)

• Grey areas
• - no fixed abode
• - short term change of residence for study /
  short term contracts with or without movement
  back and forth between origin and destination
• - ownership of more than one property
• - different definitions of migration in different
  data sources
• Model - Integration of observations from the
  various sources into a best guess synthesised
  description of underlying migration behaviour

9
Spatial Levels (chart)
10
Spatial Units (1)

• Observed changes of residence transitions over a
  given time interval very sensitive to spatial
  unit
• Purpose of migration associated with distance
• Spatial unit choice determined by reasons for
  migration
• Short distance - changes in relationships or
  accommodation
• Longer distance - education /employment/career

11
Spatial Units (2)

• Some data only fully available nationally -
  international migration outflows
• Level of breakdown of International Migration
  origin/destinations needed?
• eg. Australia as an external region?
• Pacific Island States?

12
Spatial Units (3)

• Different spatial levels as independent systems
• Hierarchy of linked migration models at different
  spatial levels - national, regional, ....
• Integrate Regional with Sub-regional Analysis?
  (eg. urban / rural, area unit and local authority
  levels)

13
Migration Data

• Census of Population and Dwellings data
• Good for five yearly transitions
• Origin, destination, but not necessarily when
  within five year period
• Years at usual address and location one year ago
  (but incomplete origin) doesn't provide
  information on multiple migration events within
  the five year period
• International Arrival / Departure Records
• Destination and TLA Origin for departures
• Origin and TLA Destination for Arrivals
• Age and Gender
• No ethnic origin data

14
Definition of the Population

• Usually Resident, Census Night Address, Overseas
  Visitors
• Higher proportion of imputed records for persons
  and households in 1996 than earlier Census's
• Higher not specified and imputed values for
  various fields in 1996 than earlier

15
Demographic Data

• Age and preferably gender
• Migration behaviour of males and females is
  different in degree at different lifecycle stages
• Need a cohort structure that is consistent with
  the migration transitions data available
• At national level one year age groups to match up
  with international arrival and departure data
  (aggregate to yearly)
• At subnational level this is at the least five
  year age groups, calibrate to 5 year group
  aggregation of national level migration model

16
Ethnic Population data

• Ethnic Composition - incomplete since no ethnic
  composition for arrival and departure data
• An Ethnic Dimension to Regional Migration Models
  is Essential
• Changes in future population are sensitive to
  shifts in future ethnic composition
• Different Age Composition of Maori, Pacific
  Island and other subpopulations
• Maori / Non-Maori Migration, Fertility and
  Survivorship Patterns
• Often Different Needs and thus Policy and
  Programme priorities for Maori, Pacific Island,
  Asian, Other subpopulations

17
Nature of the Migration Event

• Is it the individual only or is it movement of
  the family or household unit?
• Is it clearly associated with some lifecycle
  transition?
• Gross or Net Migration Flows?

18
Building the Knowledge Base

• Assembly of Data
• Adjustment / standardisation of estimates
• Calculation of Derived Values
• Integration of different sources and systems
• Association of Migration events with
  characteristics of individuals/families/households
  and spatial units

19
How to access the data

• Published tables prepared by Statistics New
  Zealand
• Customised tables analyses carried out to date
• Previous studies / projects in NZ or overseas
• Custom analyses
• Supermap X
• Supercross system analyses
• Datalab

20
Different Strategies for Sourcing Census Data
21
Application of Models / Systems

• Analysis of Broad Trends and Patterns
• Review of Theory
• Assembly of associated data for explanation
• Testing of explanatory power of theory
• Development and testing of new theory

22
Model Size

• Regional Model - 5 year cohort by gender
• 21 x 21 x 18 x 2 15,876
• Regional Model - single year age group by gender
• 21 x 21 x 101 x 2 89,082
• Regional Model - 5 year cohort by gender by
  ethnic one digit
• 21 x 21 x 2 x 18 x 6 95,256
• Local Authority Model - 5 year cohort by gender
• 77 x 77 x 18 x 2 213,444
• Local Authority Model - single year age group by
  gender
• 77 x 77 x 101 x 2 1,197,658

23
Methods for Model Extension

• Partitioning - separation of a system into
  subsystems, full interaction within subsystems
  but pooled interactions between
• Aggregation - replacement of full
  origin-destination-age-sex interactions with
  pooled migration interactions models
• Parameterisation - fitting functions to the
  migration behaviour in lieu of fully specified
  migration schedule - use in projection models
• Extension of Ethnic specific migration models
  will require use of methods for simplification
  reflecting the characteristics of each ethnic
  group