APPSIM: The Australian Dynamic Population and Policy Simulation Model

1
APPSIM The Australian DynamicPopulation and
PolicySimulation Model

• Ann Harding
• Presentation to the Technical Workshops of the
  1st General Conference of the International
  Microsimulation Association,
• Vienna, 22 August 2007

National Centre for Social and Economic
Modelling (NATSEM), University of Canberra
2
The Australian Population and Policy Simulation
Model (APPSIM)

• 5 year project, started in late 2005
• Funded by the Australian Research Council and 13
  Commonwealth agencies
• Base data is 2001 Census one per cent sample file
  (188,000 people), stored in Microsoft Access
  databse
• Parameters stored in Excel spreadsheets and
  language is C
• Full population model, with individuals being
  aged to about 2050 discrete yearly time unit

3
The overall design and flow of APPSIM
4
Processes to be modelled within APPSIM
V0.2 8 Feb 2006
5
Other characteristics

• HILDA panel data being used to estimate
  transition probabilities 5 years of data, 7000
  households, sample size problems
• Alignment required to match ABS population
  projections and Treasury Intergenerational Report
  labour force projections
• Facility to turn alignment on or off within
  each module
• User friendly interface and output envisaged
• Ability to run sub-set of sample only and write
  to output yearly, 5 yearly, 10 yearly or at end
  of simulation

6
(No Transcript)
7
(No Transcript)
8
Issues for possible discussion

• How much alignment are others doing? Is it broad
  alignment (e.g. total number of babies) or very
  detailed alignment (e.g. number of babies by
  mothers age and marital status)? How well do the
  micro equations perform without alignment? What
  is technique used to align?
• Marriage market any updates on what works and
  is efficient to run?
• Problems creating the requisite histories for
  people?
• How successful has statistical matching been?
• How have others modelled labour markets sub
  year output?
• Any hints on modelling earnings or education or
  assets?
• How have others modelled health status and what
  processes does it affect within model?