A microsimulation Analysis of the 2006 regime change in the Dutch disability benefit scheme

1
A microsimulation Analysis of the2006 regime
change in the Dutch disability benefit scheme

• Jan-Maarten van Sonsbeek
• Raymond Gradus

2
Contents

• Introduction
• Background on Dutch DI scheme
• Microsimulation model
• Small scale microsimulation model
• Used in policy making process
• Results

3
Characteristics of the (old) DI system

• High benefits (70 previous wage), even more in
  1st/2nd year of sickness ? Little incentive to
  work
• Low entrance level (15 based on income loss)
• Permanent benefit vs. temporary unemployment
  benefit
• No restrictions on number of years worked or
  cause of the disability

4
Number of disability benefits (1967-2002)
5
The Disability Insurance problem in 2002

• Almost 1 million disability benefits as compared
  to 7 million working population
• Most past policy measures failed up till then
• Ageing related policy measures ? higher labour
  participation of elderly and women
• Both elderly and women are high risk groups for
  disability benefits
• Number of DI benefits expected to rise faster
  than working population in the future

6
The structure of the 2006 DI law
7
Macro-models for DI forecasts

• Subdivide disability population in relevant
  subgroups (fully/partially disabled,
  working/unemployed, lt/gt35)
• Use of average age, wage, disability grade per
  group
• Estimate direct effect per subgroup
• Estimate behavioural effect per subgroup
• Compute average replacement rate per subgroup
• Use of benefit elasticity
• But subgroups are very heterogeneous

8
Why microsimulation for DI reform?

• Complex regime change with many benefits involved
  (3 DI benefits, UI benefit, social assistance
  etc.)
• Complex way of financing benefits from different
  funds
• Many different policy alternatives considered
• Interest in both short-term and long-term effects
• Political interest in income effects
• Since 2001 full database of beneficiaries
  available

9
Simulation model - 1

• Platform SAS
• Starting point microdatabase of DI-benefits
• Gender, Birthdate, Grade of disability, Household
  situation, Duration of benefit, Previous wage
• Extension with imputed data from other sources
• Working history, Partner income, Supplementary
  (private) insurance
• 10 samples provide sufficient accuracy
• Sample of existing cases sample of inflow
  cohorts
• Inflow exogenous
• Simulation database consists of 350.000 records

10
Simulation database
Existing cases
Source
Impute
Sim. DI
Sim. UI
Sim. SA
Sum(2004)
Sum(2040)
Inflow
11
Simulation model - 2

• Separate databases for old and new system
• Fixed random numbers in second database
• Simulation of medical examination
• Behavioural model
• Based on uniform benefit elasticity like in
  macro-forecast
• Benefit elasticity 1.5 based on international
  studies 1 DI benefit raise ? 1.5 more DI
  benefits
• Financial triggers do not influence inflow
• Financial triggers do change participation
  behaviour

12
Results Cost of benefits (2005-2040)
13
Results Average replacement rates
14
Results Overview

• Number of DI benefits - 27 (200,000)
• Cost of DI benefits - 19 ( 1.9 billion)
• Wages earned by DI population 1.7 billion
• Overall replacement rate stable
• Gini coefficient - 0.1
• Conclusion DI reform is efficient

15
Modelling developments

• Simulation model now in use for regular budget
  forecasts
• Updated source data (e.g. 2006 microdataset)
• First year of new law greater effects than
  predicted
• Yearly inflow down from 60.000 to 30.000
• Permanently fully disabled only small share of
  total inflow
• But use these with care
• Past experience
• 2007 Benefit increase permanently, fully disabled

16
Conclusions on the use of msm

• Disadvantages
• Complexity (transfer of knowledge)
• Development time
• Advantages
• Quick response time
• Consistent forecasts of different benefits
• Amount of information (a.o. both price and
  quantity)
• Convincing output