AIM-AP projects 1 and 2
Download
Report
Transcript AIM-AP projects 1 and 2
EUROMOD: the flexibility to “add new
countries”
Holly Sutherland
Institute for Social and Economic Research, University of Essex
(with Francesco Figari, Horacio Levy, Christine Lietz and Alari Paulus)
ESRC/BSPS Microsimulation seminar 2nd April 2009, LSE
2
Outline
• What is EUROMOD and what is it for?
• Context: past, present and future
• Model structure
– and the process of model construction/maintenance
• Model components
–
–
–
–
user interface
functions
data
documentation
• Challenges
3
What is EUROMOD and what is it for?
•
EUROMOD is a multi-country tax-benefit model for the European Union: unique
•
National tax-benefit models exist in most of the EU15 and some of the NMS
National models for national analysis without complex changes
EUROMOD for national analysis involving complex changes or comparative multicountry analysis or trans national research (EU level)
•
Cross-national comparisons
– equivalent outputs with different inputs
– huge range of options maximise flexibility
– defining variables that improve comparability/equivalence e.g. net social benefits, child
contingent payments, indicators of work incentives
•
The differential effects of “common” reforms
•
Policy learning across countries: “policy swapping”
─ Understanding the effects of tax-benefit systems on different populations
─ “Borrowing” policies that seem effective in one country (e.g. UK WFTC)
•
Any non-commercial research use (academic, govnt., EC, OECD etc.), free of
charge, subject to input data access permission and EUROMOD acknowledgement
4
Past and present
•
EUROMOD was originally built (for EU15) because of difficulties in making
national model calculations comparable, funded by a series of European
Commission projects (1998-2004)
•
It was first constructed very fast in a learning-by-doing way
•
Lessons learned [flexibility-comparability-ease of use and maintenance] fed into
new design applied to 4 NMS (I-CUE FP6 RIDS project 2004-8)
•
A common framework for doing equivalent things across countries
– a unified design
– common structure and building blocks
– disciplined input data specification
•
Complexity requiring
– good documentation including validation
– training and support for users
– specialist developers + national expertise
•
A further revision is ongoing now (EC DG-EMPL) to be applied to all EU27
•
The EUROMOD framework has been used for a model for South Africa and is in
process for LATINMOD (5 countries)
5
Current status
• Direct taxes and cash benefits only; no indirect taxes or noncash benefits; most contributory benefits and pensions are not
fully simulated
• Standard model: static calculations; benefit take up is assumed
to be 100%; no tax evasion.
• EU15 plus 4 NMS (EE, PL, HU, SI)
Policy
EE HU PL SI
year
1998
2001
2003
2005 + + + +
EL ES BE DE LU NL AT PT FI UK DK IE FR IT SE
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
• Data either matches policy year or precedes it a few years
(then uprated); various datasets
+
+
6
Current/future
• EUROMODupdate: DG-EMPL funding 2009-12
– extend EUROMOD to EU27
– re-base using the EU-SILC (Eurostat UDB) as the input data
– update policies to the EU-SILC data year + a very recent year
– re-write the EU-15 using the revised framework
– encourage the use of EUROMOD for policy relevant research by all relevant
people/bodies
– establish a regular (annual) programme of updating (data and policies)
– transfer as much of the maintenance as possible to national teams
– train more people to use the model
• Links to other models/developments
– indirect tax
– labour supply
– non-cash incomes
– benefit take-up and tax evasion
– new spin off models for non EU countries
7
What are the key requirements?
• All country-specific (data or policy) features should be
parameterised and not in the code
• Generic functions and variable definitions should work across
countries (“building blocks” or a common “language”)
• Transparency, and a “getting started” mode of working for
both developers and model users
Model structure
9
EUROMOD structure
Original data
Policy rules
(agreed with
national teams)
(reported by
national teams)
Model input data
Parameters
(individual-level data
prepared by national
teams,text files)
(Excel files)
Code
(written and
compiled in C++)
Micro output data
e.g., net income
(text files)
Specific analysis
(e.g., Stata, SPSS)
Tools for
basic analysis
(Excel files/macros)
10
EUROMOD parameter files
Information about the input dataset and tax-benefit system the model needs for
producing its output is stored in an Excel file (using different sheets), one set
for each country:
– CONTROL → datasets definitions and general settings
– POLICY SPINE → order of policy simulations
– POLICIES → rules of tax/benefit instruments decomposed into elements (i.e.
functions)
– TAX UNITS → definitions of assessment units
– INCOME LISTS → definitions of income concepts
– UPDATING FACTORS → to update input data monetary values to the system
year
An additional common file (Variables.xls) defines all the variables used in the
model.
Model components:
user interface (operating system)
12
Operating system
13
Running EUROMOD
Check and select countries and data-system combinations
Output path: where output files are to be saved
14
Grouping
used
Control
General
settings
Define
dataset 1
Define
dataset 2
Parameter
names
Parameter
values for
EE-2005
system
15
Policy spine (i.e. order of policies)
Function
On/off
switch
Model components:
functions
17
Policy functions (just 6 of them)
• func_Elig “eligibility function” - most frequently used for
determining the eligibility for receiving benefits; also liability
for paying taxes and evaluating other conditions
• func_BenCalc “benefit calculator”
• func_ArithOp is a simple calculator, allowing for the most
common arithmetical operations
• func_SchedCalc implementation of the most common (tax)
schedules
• func_Allocate (re)allocates amounts (incomes, benefits, taxes)
between members of assessment units
• func_Min and func_Max are simple minimum and maximum
calculators
18
Policies: a real example
19
Policies: a Simpleland example (1)
20
Policies: a Simpleland example (2)
sys_par_name
sys_reserved sys_first_sys
sys_first_par
func_Elig
elig_cond
TAX_UNIT
func_ArithOp
who_must_be_elig
formula
output_var
TAX_UNIT
sys_end_par
SL_demo
SL_reform sys_end_sys
Child Benefit
on
off
eligibility rule
{dag < 3}
{dag < 3}
in the base scenario there must be
sben_family_slsben_family_sl at least one child aged younger than three
on
on
benefit calculation
one
n/a
100#m
50#m
the benefit amounts to 100/50 € per month
bch_s
bch_s
sben_family_slsben_family_sl
• “copy” (using EUROMOD tool to add a system) and edit new columns to
create reforms to existing policies
• copy and edit policies (groups of rows) to create new policies or “swap”
policies across countries
• use the add country EUROMOD tool and copy/edit policies for new
countries
Model components:
input data and output variables
22
EUROMOD input database
• Definition and names of EUROMOD variables (input and
output) follow a standardised approach
– list of acronyms that, put together in a predetermined order, build a
variable name
• Two classes of acronyms, ordered hierarchically:
– Class 1: IDentification number, Asset, Labour market, Demographic,
System, primarY income, eXpenditure, Benefit, Pension, Taxes and
contribution, in Kind
– Class 2: two characters to identify specific information for the subject
• more than one acronym can be used
• many categories of acronym, to be used as necessary
– Simulated variables end with _s, e.g. bch and bch_s
23
EUROMOD variables example: child benefits (so far)
VARIABLES
Variable
Name
Automatic
Label
sys_var_name
sys_label
bch
bch_s
bch00
bch00_s
bchab
bchba
bchba_s
bchdi_s
bchdied_s
bched_s
bchlg_s
bchlp
bchlp_s
bchlp00_s
bchmt_s
bchnm_s
bchot
bchpl
bchunlp
benefit
benefit
benefit
benefit
benefit
benefit
benefit
benefit
benefit
benefit
benefit
benefit
benefit
benefit
benefit
benefit
benefit
benefit
benefit
:
:
:
:
:
:
:
:
:
:
:
:
:
:
:
:
:
:
:
child
child
child
child
child
child
child
child
child
child
child
child
child
child
child
child
child
child
child
:
:
:
:
:
:
:
:
:
:
:
:
:
:
:
:
:
:
simulated
main/basic
main/basic : simulated
abroad
birth/adoption
birth/adoption : simulated
disability-invalidity : simulated
disability-invalidity : education : simulated
education : simulated
large family : simulated
lone parent
lone parent : simulated
lone parent : main/basic : simulated
means-tested : simulated
not means-tested : simulated
other
parental leave
unemployment : lone parent
IT
EE
HU
child allowances -(total)
child allowance (lapsetoetus)
child allowance (lapsetoetus)
child allowance abroad
childbirth allowance
(sünnitoetus
childbirth allowance
(sünnitoetus
school allowance- (koolitoetus)
large family allowance
(kolme- ja
single parent child
- allowance (ük
regular child pro
family allowance
Child benefit (Assegno
per famiglia
- con almeno 3
-
Model components:
documentation
User guides
Developer guides
Country reports
Recipes and research applications
See http://www.iser.essex.ac.uk/research/euromod
25
Challenges
•
•
•
•
•
Input data access, quality, comparability and suitability
Enlarging and nurturing the user community
Managing model revisions and transitions
IP issues related to spin-off models
Maintaining the model developer team
– we are recruiting (closing date 27th April): see http://jobs.essex.ac.uk or
email [email protected]
26
27
Definition of income concepts
28
Definitions of assessment units
29
Uprating factors
30
Variable definitions