Introduction to STINMOD and Microsimulation Modelling in Australia

advertisement
Introduction to STINMOD and
Microsimulation Modelling in
Australia
Ben Phillips: Principal Research Fellow, NATSEM, 21 Feb 2015
What are microsimulation models?
● Tools to simulate government programs, demographic
and economic change, current or alternative situation
● Based on microdata
●
Records of individual people or households
●
Usually large – thousands of records
●
Sample surveys or administrative data
● Allow detailed assessment of impact of change
●
On individuals, or groups of individuals
●
On whole population
●
On government budgets
2
What are microsimulation models?
● Used widely in USA, UK, Europe in policy development
● Started in Australia in mid-1980s
● Types of microsimulation models
●
Static – day after impacts (STINMOD)
●
Behavioral – incorporates modelled behaviour change (MITTS)
●
Dynamic – the future (DYNAMOD, APPSIM)
●
Spatial – small areas (SpatialMSM)
3
Why microsimulation models?
●
Every year, government introduces various changes to
tax and transfer payment system
●
Goods and services tax (2000)
●
Welfare to work package (2006)
●
Major cuts in income tax (2007, 2008, 2010)
●
Increases in pension (2009)
●
Carbon Price, lower personal income tax and higher benefits
(2012)
●
Budget 2014-15 (examples)
●
Question: How do we ensure that the tax and transfer
payment system is fair, or achieve the intended
targets?
●
The impact of these policies can be assessed by using
microsimulation models on cost/revenue/distributional
outcomes.
4
Overview of STINMOD
● STINMOD (Static Income Model) is a static
microsimulation model of Australian income tax and
social security systems
● First release STINMOD 94; updated and released
every year in 3 versions (source code, outyears, and
Interface)
● Shows impact of policy changes
●
Fiscal (revenue and expenditure)
●
Distributional (winners and losers)
●
Effective marginal tax rates
5
Reduce tax for low income, Increase tax for
high income
● First rate reduced from 19% to 15% and top rate up
from 45% to 50%
6
Who wins and loses?
7
STINMOD basefile
● Created from ABS microdata (Survey of Income and
Housing – SIH and ABS Census sample file)
● Contains demographic and income information for
members of the family
●
Age, sex, marital status, study status, employment status,
housing status
●
Family type, number of children
●
Income from private sources and the government
8
STINMOD basefile
How do we make survey data representative of
the population we want to model?
● Income uprating
●
All monetary variables need to be uprated
● Ageing / reweighting
●
Adjust weights to take account of changes in the
population
●
Adjust weights to reflect number of people receiving
various social security payments
● STINMOD 14 outyears: covers 2013-14 to 2017-18
● Capable of modelling out to 2084
9
STINMOD parameters datasets
● Parameter datasets
●
Centrelink: pension rates, allowances, rent assistance,
family assistance
●
DVA: pension rates and rent assistance
●
Medicare parameters
●
Tax parameters: tax scale, tax offset, HECs repayments
●
Childcare parameters: child care benefit, childcare tax rebate
● Parameters for future years are projected based on
projected CPI, AWE (Average Weekly Earnings)
10
STINMOD program modules
● Social security payments
●
Age Pension
●
Disability Support Pension
●
Carer Payment
●
Wife Pension
●
Parenting Payment Single
●
Pensioner Education Supplement
●
Pharmaceutical Allowance
●
Rent Assistance
●
Widow Allowance
11
STINMOD program modules
● Social security payments (continued)
●
Newstart Allowance
●
Mature Age Allowance
●
Youth Allowance
●
Parenting Payment Partnered
●
Partner Allowance
●
Sickness Allowance
●
Special Benefit
●
Telephone and utilities allowance
12
STINMOD program modules
● Family payments
●
Family Tax Benefit Parts A and B
●
Maternity Payment
●
Child Care Benefit and Childcare Rebate
●
Veterans’ payments
●
Service pension (age and invalidity)
●
DVA disability pension
●
War widow(er)’s pension
●
Income support supplement
●
DVA rent assistance
13
STINMOD program modules
● Taxation
●
Income tax
●
Medicare levy
●
More common tax offsets - low income, Senior
Australian, pensioner/beneficiary, dependent spouse
●
Imputation credits
●
Child care tax rebate (hypothetical only)
●
HECS repayments (hypothetical only)
●
Education tax refund
●
Now also included carbon tax, GST, production taxes,
in-kind benefits (modelled)
14
STINMOD Output: base and new
● Simulation output include all original variables in the
basefiles and many new variables related to:
●
Entitlement to: Allowances; Pensions; Family assistance; Child
Care Benefit (hypothetical only)
●
Liability for: Income tax; Medicare levy; HECS repayments
(hypothetical only)
● From simulation output, various tables can be
produced
● Gains or losses from policy changes between base
and new simulation can be calculated
15
Examples of uses of STINMOD
● Used by Federal Government departments for
formulating budget policy
●
Federal Treasury (Capita model – developed from
STINMOD)
●
Department of Social Services
●
Department of Employment and Education
●
Recent use example: Introducing Carbon Price package in
Australia, Budget 2014-15.
●
http://www.cleanenergyfuture.gov.au/wpcontent/uploads/2012/06/CleanEnergyPlan-201206283.pdf
●
Page 37 to 51 – who wins, who loses.
●
STINMOD/microsim modelling is best when designing a
‘package’ of reform.
16
Budget 2014-15 – distributional impact
17
Hypothetical Example
18
● Introduction of GST in 2000
● Carbon Pricing in 2012
● Welfare to work reforms 2006
● Works best for costing programs where there are ‘new’
recipients or a package of policy changes are being
made.
● Adding behavioural element is complicated but useful
for describing relativities in magnitude and direction of
policy outcomes.
19
20
21
2% increase in mortgage rates – SA3
Summary
● STINMOD is complex model of Australia’s tax and
transfer system.
● Heavily used by major government departments for
policy modelling and policy costing.
● Most valuable for analysis of ‘distributional outcomes’ of
policy and considering impacts of a ‘package’ of reform.
● These types of models became popular as more
individual survey and administration data became
available in 80s and computing power increased.
● Valuable tool for policy modelling – indispensible in
Australia now.
23
www.natsem.canberra.edu.au
Download