Transcript Issues in the Use of Equivalence Scales - comments

Issues in the Use of
Equivalence Scales
- comments
Lars Osberg
Economics Department,
Dalhousie University
Workshop on Low Income, Poverty and Deprivation
Statistics Canada, Ottawa
February 12, 2007
When might the choice of
equivalence scale matter?
Percentage distribution of households,
by number in household
Number in
Household
1
All Canada
Households >65
Households <65
Single Parent Hhlds
Immigrant Head
2
3
4
5
6
7
8
How much does choice of equivalence scale
matter for empirically relevant family sizes?
Number of Equivalent Adults
according to Family Size
equivalence scales
family
LIS
Statistics
OECD
size
β = 0.5
Canada
2 adults+(n2)kids
β = 0.3
β=
1
β = 0.7
1
1
1
1
1
1
1
2
1.4
1.4
1.7
1.2
1.6
2
3
1.7
1.7
2.2
1.4
2.2
3
4
2
2
2.7
1.5
2.6
4
5
2.2
2.3
3.2
1.6
3.1
5
6
2.4
2.6
3.7
1.7
3.5
6
7
2.6
2.9
4.2
1.8
3.9
7
8
2.8
3.2
4.7
1.9
4.3
8
9
3
3.5
5.2
1.9
4.7
9
10
3.2
3.8
5.7
2
5
10
Equivale nt Adults for hous e hold s ize
12
10
Equivalent Adults
8
LIS β = 0.5
Statistics Canada
OECD 2 adults+(n-2)kids
OECD 2 adults+(n-2)kids
6
exponential β= .1
exponential β = 0.3
exponential β = 0.7
exponential β = 1
4
2
0
1
2
3
4
5
6
7
Num be r in Hous e hold
8
9
10
SST Index for Persons with Age<65 by Equivalence Scale Parameter
Luxembourg(2000)
Finland(2000)
0.13
R.O.C.Taiwan(1995)inth
US
0.12
Norway(2000)
France(1994)
0.11
IT
Sweden(2000)
0.10
IS
0.09
Denmark(1992)inth
CN
AS
0.08
SST Index
Belgium(1997)inth
Switzerland(1992)inth
0.07
NL
0.06
SP
UK
0.05
IE
BE
SW
0.04
0.03
FI
Austria(1997)inth
AT
FR
Ireland(1996)
Spain(1990)inth
DK
United Kingdom(1999)
NW
LX
0.02
CH
Netherlands(1994)inth
RC
Israel(1997)
0.01
Canada(1998)
0.00
0.000
Italy(2000)
0.100
0.200
0.300
0.400
0.500
0.600
0.700
0.800
Equivalence Scale Parameter
0.900
1.000
1.100
1.200
Australia(1994)
United States(2000)
Why is it only size that matters?

Poverty index P defined over frequency distribution of
resources R and household characteristics C

P = p( f ( R*))


P = p( f (well-being))



R* = effective resources | characteristics = r (C)
u = u (R, C)
In principle, many characteristics may matter
Household characteristics C = X + Y + Z

X – ethically irrelevant


Y – empirically irrelevant for measurement of level of poverty


Do not matter for measurement because do not matter for policy
May be highly relevant for program delivery
Z – relevant for poverty measurement

N = number household members not only relevant characteristic
Ethically irrelevant characteristics

Not all characteristics that may be empirically
important as determinants of a household’s cost
of living should be considered in poverty
measurement
 Policy

implications unpalatable
Examples:
 Gender
 Clothing costs inequality?
 Religion
 Dietary rules? Cost of private schooling? Cost of religious
prohibitions on particular activities?
Empirically irrelevant
characteristics


Not all empirically important determinants of cost
of living need to be considered in poverty
measurement
Example:
 Gender mix of children - huge impact on cost
 HRDC Market Basket for 2 Adult 2 Child family


of living
2.5 Bedroom Apartment – 50% probability of same sex
Issue: E ( p( f( R, Y, Z))) = E ( p (f ( R, Z)))
 Is
Y randomly distributed w.r.t Z ?
 Type 1 = Type 2 error

OK for Statistics Canada to ignore Y but program
administrators need to know
More than size matters!
– for measurement and analysis

Current LICO measurement practice

Number of household members


Size of urban area



Empirically important as both



Intended as adjustment for price level of needs
But if ethically relevant needs vary by other household
characteristics (Z) – analysis will mislead
Disability


Intended as adjustment for quantity of needs
cost of living determinant – wheelchairs cost $$$
policy analysis variable – increase probability low income
Age of children ? Frail elderly ?
Survey evidence needed on Canadians views on
ethically appropriate differentials in household need