Rissa Ota

1
A new New Zealandstatic microsimulation model
challenges with data

• Rissa Ota
• Helen Stott

2
Presenting Today

• Introduction
• The new model
• Survey of Family Income and Employment (SoFIE)
• Challenges
• Imputation
• Benchmark and Calibration
• Quality Assurance

3
Introduction

• What do we want to achieve
• Database for population based microsimulation
• Focused on beneficiaries
• Able to extend policies outside the cureent
  benefit system
• As small groups as possible

4
Calibration Process
SoFIE data
Imputations
Synchronisation and indexation for base and out
years
Indices
Benefit rates
Model calculation for base and out years
Benchmarks
Calibrated data
Calibration to base and out years
5
Survey of Family Income and Employment (SoFIE)

• Longitudinal survey (8 years)
• Rich income and family spells data
• Wave 1 (2002/03)
• Benefit numbers under-represented
• Short benefit spells under-represented

6
Challenges

• Data from 2002/3
• One form of support phased out
• New tax credits
• Decrease in unemployment benefit
• Inconsistencies with benefit rules
• Missing data
• Multiple potential calibration variables

7
Imputations

• Simplicity is the ultimate sophistication
• Leonardo da Vinci
• Benefit subtypes
• Beneficiaries partner
• Family spell
• Housing New Zealand
• Boarding
• Childcare use
• Benefit eligibility in each income/family spell

8
Benchmarking
9
Sharpness Too High
Correct Reproduction
Gamma Too High
10
Calibration

• From 2002/03 to 2004/05
• Person Level
• Age by sex
• Labour force status
• Born in NZ
• Household or Family Level
• Household composition
• Deprivation index
• Benefit
• Benefit spell period by benefit type
• Number of people and families by benefit type
• Benefit expenditure

11
Relative Std Error of some statistics
12
Further Developments

• More robust testing
• More data - new waves
• Extending model to other parts of the benefit
  system
• Alignment with administrative data
  microsimulation model