SimYorkSimBritain: a spatial microsimulation approach to modelling spacetime population dynamic proc

1
SimYork/SimBritain a spatial microsimulation
approach to modelling space-time population
dynamic processes
Geoinformatics research group seminar, School of
Geography, University of Leeds, Thursday 14
February 2002
Dimitris Ballas, Graham Clarke, Danny Dorling,
Heather Eyre David Rossiter
2
Outline

• What is microsimulation
• Spatial microsimulation approaches
• The SimYork/SimBritain model data and methods
• Conclusions

3
What is microsimulation?

• A technique aiming at building large scale data
  sets
• Modelling at the microscale
• A means of modelling real life events by
  simulating the characteristics and actions of the
  individual units that make up the system where
  the events occur

4
What is spatial microsimulation? An example

• sex by age by economic position (1991 UK Census
  SAS table 08)
• level of qualifications by sex (1991 UK Census
  SAS table 84)
• socio-economic group by economic position (1991
  UK Census SAS table 92)

5
Spatial microsimulation procedures

• Construction of small area microdata from using
  samples, surveys and small area data
• Static-what-if simulations
• Dynamic modelling to update the static microdata
  set and perform dynamic what-if micro-spatial
  policy analysis

6
Spatial microsimulation methodologies

• Probabilistic synthetic reconstruction techniques
  (IPF-based approaches)
• Combinatorial optimisation methods
  (hill-climbing, simulated annealing, genetic
  algorithms)
• Event modelling

7
Simulating migration, education and social
mobility
It is well known that mobility rates are
substantially higher among renters than among
homeowners. Similarly, the age structure of
migrants to and from neighborhoods is likely to
be quite different in a neighborhood comprised
primarily of homeowners in comparison with a
renter-dominated neighborhood.
(Rogerson and Plane, 1998 1468)
During their lifetimes, the simulated
individuals have to change their educational and
employment status. They will enter school with
different probabilities when they are between 14
and 20 years old, they will be employed in
different jobs, lose their jobs, earn an income
which depends on their type of job, and
eventually retire with different probabilities
depending on their ages.
(Gilbert and Troitzch, 1998)
8
SimYork dynamic spatial microsimulation
9
Data sources Census data and the BHPS

• 1991 Census of UK population
• 100 coverage
• fine geographical detail
• Small area data available only in tabular format
  with limited variables to preserve
  confidentiality
• cross-sectional

• British Household Panel Survey
• sample size more than 5,000 households
• Annual surveys (waves) since 1991
• Coarse geography
• Household attrition

10
Dynamic simulation data sources

• National Health Service Central Register (NHSCR)
• -internal migration data by single year of age
  and sex
• -for 98 FHSAs in England, 8 in Wales, 5 in
  Scotland (now available at district level)
• 1991 Census, Special Migration Statistics
• - down to ward level
• - 5-year age groups by sex

11
SimYork/SimBritain aims and objectives

• Reweight the first wave of the BHPS data to fit
  electoral wards
• Dynamically simulate this population for each
  year up to 2001
• Dynamic simulation of York/Britain to 2011/2021
  (groundhog day scenario)
• What-if dynamic simulations

12
(No Transcript)
13
Census Small Area Statistics (SAS) tables to be
used as small area constraints

• Tenure by students in households (SAS table 26)
• Infrastructure rooms by dwelling type (SAS table
  57)
• Economy earners by number of dependent children
  (SAS table 36)
• Demography age, sex and marital status of
  household head (SAS table 39)

14
Reweighting the BHPS

• Find the combination of BHPS households that best
  fit small area statistics tables
• simulated annealing
• hill climbing
• genetic algorithms
• new methods!

15
Reweighting the BHPS - a simple example (1)
A hypothetical sample of individuals (list format)
Hypothetical Census data for a small area
In tabular format
16
Reweighting the BHPS - a simple example (2)
Calculating a new weight, so that the sample
will fit into the Census table
Hypothetical Census data for a small area
In tabular format
17
Simulating Clifton
18
Reweighting the BHPS - a real example SimClifton
Using the BHPS to reproduce part of Census table
57
Small Area Statistics data from table 57, Clifton
ward, York, 1991
19
Reweighting the BHPS - a real example SimClifton
Using the BHPS to reproduce part of Census table
39
Small Area Statistics data from table 39, Clifton
ward, York, 1991
20
Reweighting the BHPS - a real example SimClifton

• A deterministic approach
• Reweight the BHPS to fit table 39 population
  distribution for Clifton
• Reweight the BHPS to fit table 57 population
  distribution for Clifton
• Reweight the BHPS to fit table 26 population
  distribution for Clifton
• Reweight the BHPS to fit table 36 population
  distribution for Clifton

21
Results from reweighting the BHPS to fit tables
39 and 57 for Clifton
Table 39 (reproduced using SimClifton output)
Table 57 (reproduced using SimClifton output)
22
Total absolute error
After 100 iterations
After 5,000 iterations
23
Integer weighting
24
Integer weighting

• 1. Set a rounding variable equal to 0.999
• 2. Increase by 1 the weight of all households
  that have a decimal remainder bigger than or
  equal to rounding
• 3. Decrease rounding by 0.0001
• 4. If sum of all weights is equal or bigger than
  total number of households in the area then exit.
• 5. Return to step 1

• 1. Set a cp and an i variable equal to zero
• 2. Read the array of sorted households hi (i
  0 - total number of BHPS households)
• 3. Increase cp by the weight of the next sorted
  household h(i)
• 4. If cp 1 give to the household h(i) an
  integer weight equal to the rounded cp value and
  subtract this value from cp (e.g. if cp 2.05
  set household weight 2 and set cp 2.05 -2
  0.05). Increase i by 1 (move to next household)
• 5. If ielse exit

25
Geographical weighting (1)
26
Geographical weighting (2)
27
The potential of SimYork/SimBritain for policy
analysis
Source The Guardian, 22 March 2000
28
The potential of SimYork/SimBritain for policy
analysis
Academic jokers
Source Leeds School of Geography web-site
29
Using SimYork for policy analysis - estimating
non-census attributes
30
Using SimYork for policy analysis - modelling
electoral behaviour
31
Using SimYork for policy analysis - modelling
electoral behaviour
32
Conclusions

• Build on existing spatial and aspatial
  microsimulation models
• Policy spatial micro-modelling - income and
  substitution effect
• Include more regional subsystems (labour demand,
  schools, hospitals, etc.)
• Small area multiplier analysis
• What-if, what-will-happen-if and
  What-would-have-happened-if analysis